solving the ice age mystery: the deep-ocean solution

12Solving the Ice Age Mystery:

The Deep-Ocean Solution

12.1 Astronomical Drivers

By the time of the Pleistocene, the Earth had cooled to itslowest point since the Carboniferous glaciation 300 Maago. During the Pleistocene, much of the variation in theEarth’s climate was due not to the plate tectonic pro-cesses that were important in older times, but to celestialmechanics. This led to cycles of 100 Ka in the eccentricityof the Earth’s orbit, of 41 Ka in the Earth’s axial tilt (whichdominates radiation at high latitudes) and of 22 Ka inthe precession of the equinoxes (which dominates radia-tion at low latitudes)1. As we saw in Chapter 6, MilutinMilankovitch set out the basis for our understanding ofthis process between 1920 and 1941.

In 1945, Frederick Zeuner (1905–1963) of London’sInstitute of Archaeology tested Milankovitch’s theory byexamining how it applied to what was known of the Pleis-tocene period on land2. Finding a close match between thesequence of Ice Age strata and the variations in insolation,he concluded, ‘no objection can be raised against theastronomical theory of the glacial and interglacial phasesof the Pleistocene’2. Among those intrigued by whatcontrolled the changes between glacial and interglacialperiods was Richard Foster Flint (1901–1976) of YaleUniversity, who, along with other honours, would beawarded the Prestwich Medal by the Geological Societyof London in 1972 for his contributions to our understand-ing of the Ice Age. Flint was one of the most influentialfigures in Quaternary science in the 20th century, muchadmired for his seminal 1957 text Glacial and Pleistocene

Earth’s Climate Evolution, First Edition. Colin P. Summerhayes.© 2015 John Wiley & Sons, Ltd. Published 2015 by John Wiley & Sons, Ltd.

Geology3. This book concluded, ‘the geometric schemeof distribution of insolation heating must be consideredinadequate in itself to explain the Pleistocene climaticchanges’3.

Things have changed since then. As Mike Walker of theUniversity of Wales and John Lowe of Royal HollowayCollege London pointed out in 2007, Flint’s approach wasrooted in glacial geology. Since his day, those investigat-ing Quaternary science have moved ‘away from Flint’ssomewhat narrow glacial-geological paradigm towardsthe multi- and inter-disciplinary approach to the study ofrecent Earth history that is practiced today’3. With thisnew approach, we can now analyse the ‘rich and oftenreadily accessible Quaternary record… at a level of detailnot normally possible for older geological periods’3 .Milankovitch’s theory has come to stay4.

Milankovitch lacked computers. André Léon GeorgesChevalier Berger (1942–) (Box 12.1) used them to refinehis calculations5–8. Figure 12.1 provides an introductionto Berger’s findings, which we explore in more detail inChapter 13.

Full comprehension of Ice Age climate change hingeson novel studies of deep-ocean sediments collectedby piston cores and deep-ocean drilling, as we see inthis chapter, and of ice cores, which we examine inChapter 13. These studies fall into the science of theQuaternary, which comprises the Pleistocene, starting at2.6 Ma ago, and the Holocene – the last 11 700 years3. Tohelp move the field forward, scientists formed the Inter-national Quaternary Union (INQUA) in 19283. Several

Solving the Ice Age Mystery: The Deep-Ocean Solution 241

eccentricity (e)

climatic precession (e sin ϖ)

0.05

0.06

0.04

0.03

0.02

0.01

0.06

25

24

23

22

100 50

obliquity (ε)

insolation 65N June

0 –50 –100 –150Time (kyr)

–200 –250 –300 –350 –400

450

500

–0.06

–0.04

–0.02

0.00

0.02

0.04

Figure 12.1 The Berger Astronomical Model of Orbital Variability Present and Future. These curves have beenproduced in numerous formats in several publications by André Berger and Marie-France Loutre. Obliquity isexpressed in degrees of tilt of the Earth’s axis. Insolation at the summer solstice at 65∘ N is expressed in W/m2.

Box 12.1 André Léon GeorgesChevalier Berger.

André Berger has a master’s degree in meteorol-ogy from MIT (1971) and a doctorate from theCatholic University of Louvain, Belgium (1973).He is renowned for contributing to the renaissanceof Milankovitch’s theory of climate change, formaking major contributions to simulating futureclimate change and for working on the first Earthmodel of intermediate complexity. He was profes-sor of meteorology and climatology at Louvain,and then director of the Institute of Astronomy and

Geophysics Georges Lemaı̂tre from 1978 to 2001,where he now has emeritus status. He has servedas president or chairman of several national andinternational scientific organisations and commit-tees and was honorary president of the EuropeanGeosciences Union. He was on the steering com-mittee for the International Geosphere-BiosphereProgramme (IGBP) and initiated the PalaeoclimateModeling Intercomparison Project (PMIP). Hehas received many honours for his discoveries,including the Milutin Milankovitch Medal of theEuropean Geophysical Society, and in 1996 he wasmade a knight of the realm by King Albert II.

242 Earth’s Climate Evolution

scientific journals emerged to meet their needs: QuaternaryResearch (1971), Boreas (1972), Quaternary ScienceReviews (1982), Journal of Quaternary Science (1985),Quaternary International (1989) and The Holocene(1991). Researchers also make good use of journalslike Paleoceanography and Palaeogeography, Palaeoe-cology, Palaeoclimatology, as well as the four-volumeEncyclopaedia of Quaternary Science (2006)9.

12.2 An Ice Age Climate Signal Emergesfrom the Deep Ocean

As we saw in Chapter 7, the first to exploit the new tech-nology of piston coring in order to examine the historyof climate recorded in deep-sea sediments was GustafArrhenius. In 1952, Arrhenius attributed alternationsbetween carbonate-rich and carbonate-poor sedimentsin east Pacific cores to changes in the ‘aggressiveness’of polar bottom waters. During the Ice Age, he thought,large volumes of bottom water were derived from thepolar regions. Being cold, they carried large amounts ofdissolved CO2, which enabled them to dissolve, or toprevent the deposition of, deep-sea carbonates. Bottomwaters of intervening warm periods carried less dissolvedCO2, so were less ‘aggressive’. At the time, Arrhenius,like Flint, dismissed Milankovitch’s ideas10, although helater adopted them.

At about the same time, in the early 1950s, Lamontbegan its routine collection of long piston cores fromthe world’s oceans. David Ericson, who was in chargeof the new Lamont core store, found that although thedistribution of the planktonic foraminiferan Globorotaliamenardii indicated warm conditions, it was also influencedby ocean currents. Another species, Globigerina pachy-derma, was a cold-water indicator, as were Globigerinainflata and Globigerina bulloides. Changes from warmto cold were also indicated by changes in the coilingdirection of Globorotalia truncatulinoides. From thedistribution of these species down-core, Ericson built aQuaternary stratigraphy incorporating the Holocene, thelast glaciation and the previous interglacial11.

Ericson and Arrhenius’s interests were shared by CesarEmiliani, whom we met in Chapters 6 and 7. In 1955,Emiliani analysed the oxygen isotopes in planktonicforaminifera collected from eight piston cores from theSwedish Deep-Sea Expedition and four from the Lamontcore store. Finding fluctuations in the 𝜕

18O ratio withtime, he interpreted them as representing the variations

in climate between glacial and interglacial periods12.By analysing the same species of surface-dwellingforaminiferan, he eliminated the effect of metabolic differ-ences between different species. The resulting variationswere due to differences in either the temperature or the iso-topic composition of seawater, the latter representing theamount of water tied up as ice on land. Emiliani ‘guessedthat 60% of the signal was due to the temperature effect,40% to the ice effect’13. Following Zeuner’s reasoning2, hethought that the variations in 𝜕

18O with time representedchanging insolation12. He invented the nomenclature thatis still in use today, in which each warm or cold periodis identified as a marine isotope stage (MIS). MISs witheven numbers are cold stages, those with odd numbers arewarm stages. Some can be subdivided into substages (e.g.5a, 5b, 5c).

In March 1961, Emiliani teamed up with Flint tolink the Pleistocene record in continental and deep-seasediments14. They considered that MIS 1 was theHolocene, MIS 2 was the main Würm glacial stageon land, MIS 3 was the early to main Würm glacialinterval on land, MIS 4 was the early Würm on landand MIS 5 was the last interglacial. Thinking about thepossible drivers for the Ice Age, they favoured a modelof glaciation based on Milankovitch’s concept that iceaccumulated during cool summers, driven by insolation inthe Northern Hemisphere. Evidently, Flint had changedhis mind about Milankovitch since 1957. They discountedPlass’s idea that ice ages were in some way controlledby atmospheric CO2, because they thought, wrongly, thatRevelle and Suess15 had concluded, based on studies of14C, that atmospheric CO2 would be rapidly taken up bythe ocean. It would be a while before the role of CO2 inthe Ice Age climate would be fully understood.

Before Emiliani’s isotopic analyses of deep-sea cores, itwas thought that there were only four major glacial peri-ods during the Ice Age2. Emiliani found more than twicethat! While this shocked classical Quaternary geologists,physicist Nicholas J. Shackleton (Box 7.3) agreed withEmiliani16. The fly in the ointment was the inadequacy ofmethods for accurately dating Emiliani’s cores. Emilianiguessed that his main 𝜕

18O cycles were 41 Ka long, butthey were later found to last about 100 Ka.

Subtle differences in the ways in which Ericson andEmiliani interpreted their data meant that, for a while, theydisagreed about the precise sequence of cold and warmevents. When they did agree, we had two independentmeans of mapping changes between glacial and inter-glacial periods in deep-sea cores – one palaeontological,


Figure 12.2 John Imbrie.

from microscopic fossils of marine plankton (microfos-sils), and the other geochemical, from oxygen isotopes inthose same fossils.

By the mid 1960s, two new figures occupied centrestage: Nick Shackleton and the micropalaeontologist andstratigrapher John Imbrie (1925–), of Brown University(Figure 12.2, Box 12.2). Our understanding of Ice Ageclimate owes a great deal to these two men.

Box 12.2 John Imbrie.

Imbrie obtained a BA from Princeton in 1948,after serving with the US 10th Mountain Divisionin Italy during the Second World War. Havingobtained a PhD from Yale in 1951, he taught atColumbia University until 1967. Joining BrownUniversity, he held the Henry L. Doherty Chair inOceanography, and he now holds emeritus statusthere. He pioneered the use of computers to demon-strate the relation of assemblages of plankton tothe temperature of surface waters, thus providingpalaeoceanography with one of its key tools. Hisbook Ice Ages: Solving the Mystery18, written withhis daughter Katherine, won the 1976 Phi BetaKappa Prize. Imbrie was co-author with Hays andShackleton of the 1976 paper in Science that linkedMilankovitch variations to the sediment record.He was elected to the US National Academy ofSciences in 1978 and received the Maurice EwingMedal of the American Geophysical Union in 1986,

the Twenhofel Medal of the Society of SedimentaryGeology, the Lyell Medal of the Geological Societyof London in 1991 and the Vetlesen Prize in 1996.

Using his refined mass-spectrometric technique (seeChapter 7), Shackleton showed by 1967 that a distinctive𝜕

18O pattern is evident in the benthic foraminifera that liveon the seabed, which are bathed in cold oxygen-rich bot-tom water sinking from the surface in the polar regions17.Most of this water is colder than 4 ∘C, so short-termvariations in the 𝜕18O ratio in these organisms tell us moreabout changing seawater temperature than changing icevolume. Shackleton found that planktonic foraminiferagrowing in surface waters displayed much the same signal.He calculated that about 66% of the 𝜕

18O shift betweenglacial and interglacial periods was due to changes inice volume, not to the influence of temperature, whichwas the opposite of what Emiliani had concluded. Thisrevelation caused a paradigm shift in our understanding of𝜕

18O ratios in the service of palaeothermometry. Urey hadassumed that the oxygen isotopic composition of seawaterwould be invariant19. Clearly it was not. Today, the 𝜕

18Oratios of benthic foraminifera are taken as representing thetemperatures of polar surface waters.

By 1969, Imbrie realised that because the total assem-blage of planktonic foraminifera in surface waters shouldreflect the environment in which they lived, and multivari-ate statistical analyses could quantify that relationship, sta-tistical analyses of faunal assemblages down-core could beused to ascertain past climate change20. Reanalysing theCaribbean cores analysed by Ericson and Emiliani, he andNilva Kipp showed that the temperatures of surface watersof glacial periods there fell by just 2 ∘C, not 6 ∘C. Becausethe fluctuations in their data agreed with those determinedby Emiliani from isotopes, it was clear that Globorotaliamenardii, which Ericson used to identify warm periods,fluctuated in ways unrelated to temperature, explaining thediscrepancy between Ericson and Emiliani’s data21, 22. Thiswas a breakthrough.

By the late 1960s, as the Deep Sea Drilling Project(DSDP) got underway, analyses of 𝜕

18O changes downpiston cores from different parts of the ocean showed thatsediments could be routinely subdivided into Emiliani’sMISs representing glacial and interglacial periods. Thesestages coincided with intervals defined by Imbrie’s assem-blages of microfossils and could be correlated from onecore to another over vast oceanic distances, suggestingplanetary control.


Was that planetary control the same as Milankovitch’sastronomically controlled insolation? To find out, geol-ogists needed closely spaced dates down-core. In thelate 1960s to early 1970s, microfossils could tell usabout environmental change from cold to warm andback, but not about the ages of cold and warm stages.As we saw in Chapter 6, radiocarbon dating was useful,but only in sediments less than 50 Ka old. Layers ofvolcanic ash older than 100 Ka could be dated by thepotassium–argon (K-Ar) method. Changes in the Earth’smagnetic field could be used to date specific sedimen-tary horizons, but only in sediments older than 780 Ka.Together, these independent techniques provided a crudemeans of dating sediment layers in cores. Over time,more techniques would become available. A major break-through in radiocarbon (14C) dating came about in 1977,when a new technique – accelerator mass spectrometry(AMS) – enabled us to count 14C atoms, as opposed tomeasuring 14C decay. AMS can date samples as smallas a pinhead-sized microfossil. The technique is fast andcheap, but it is still limited to sediments less than about50 Ka old.

Determining the ages of the glacial and interglacialsedimentary stages identified by 𝜕

18O and microfossilanalyses became a major objective of the internationalClimate Long Range Investigation, Mapping and Predic-tion (CLIMAP) project. Founded by Imbrie, Shackletonand others, CLIMAP began in spring 1971 as part ofthe International Decade of Ocean Exploration23. Theproject aimed to establish average boundary conditionsfor the Last Glacial Maximum at 18 Ka ago. Those con-ditions included the geography of the continents, thealbedo of land and ice surfaces, the extent and elevationof permanent ice and the sea surface temperature. Mod-ellers would use those conditions in atmospheric GeneralCirculation Models (GCMs) to map the climate of theLast Glacial Maximum. CLIMAP would then test modeloutputs against palaeoclimate data. The first simulationwas for August 18 Ka ago23. Sea surface temperaturevalues were derived from 𝜕

18O data and from Imbrie’sstatistical analyses of planktonic faunal assemblages. Theextent of sea ice in the polar regions was estimated fromthe presence or absence of diatomaceous sediments, withabsence indicating ice.

The CLIMAP data showed that extensive cooling at thepoles and an expanded area of land and sea ice steepenedthe thermal gradient between the Equator and the poles,strengthening the winds. In the Southern Ocean, theAntarctic polar front moved north, along with Antarctic

sea ice. The Subtropical Front moved far enough northto limit the passage of warm Indian Ocean water aroundSouth Africa and into the South Atlantic. This cooled theSouth Atlantic and created a closed anticlockwise gyre inthe Indian Ocean. Upwelling increased where it is foundtoday, along continental margins and along the Equator,as the winds that drove it increased in strength. On land,grasslands, steppes, deserts and ice spread at the expenseof forests, increasing the Earth’s albedo.

Palaeoclimatologists noticed that the wiggles in the 𝜕18Ocurves down sediment cores seemed to mimic the wigglesin the patterns of Earth’s insolation through time. If thematch was real then it offered an opportunity to date the ageof the sediments from the pattern of wiggles in the oxygenisotope data. Starting with just a few radiometric dates astie points, the CLIMAP scientists assumed that the rates ofsedimentation in different MISs were constant down-core.This enabled them to estimate the age of each wiggle onthe curve of variation in 𝜕

18O back through time. Applyingspectral analysis to the 𝜕18O curve dated in this way, JamesHays (1938–) of Lamont, together with Imbrie and Shack-leton, demonstrated in a landmark paper in 1976 that incores where sedimentation was undisturbed, the variationsin 𝜕

18O over the past 450 Ka accurately mimicked theorbital signals calculated by André Berger24. Furthermore,climate changes in the Northern Hemisphere were essen-tially synchronous with those observed in the SouthernHemisphere. The correlation between the astronomicalvariables and 𝜕

18O told them that ‘changes in the earth’sorbital geometry are the fundamental cause of the succes-sion of Quaternary Ice Ages’24. The three Milankovitchmechanisms – eccentricity, tilt and precession – worked inunison to provide the ‘pacemaker of the ice ages’.

Not only that, but the fact that the same pattern of wig-gles occurred everywhere meant that even cores withoutprecise radiometric dates could be dated by reference to astandard 𝜕

18O curve derived by merging data from severalwell-dated cores. Wiggle matching offered an incredibleopportunity to date intervals of time as small as about1000 years long. Patterns of tree rings back through timeoffer much the same possibility, provided they come frommuch the same area and so experienced more or less thesame climate changes through time. While this is veryhelpful for the past 11 Ka, tree rings do not provide uswith a lengthy and globally distributed data base of thekind provided by deep-sea cores. Wiggle matching ofoxygen isotope curves from core to core does containthe assumption that the section has not been disturbedby either erosion or the lateral introduction of material


by turbidity currents, but these possibilities are identifiedor eliminated by comparing individual cores with theglobal standard. Indeed, wiggle matching can identifyhow much section has been removed! Nowadays, wigglematching has enabled palaeoclimatologists to push the𝜕

18O-calibrated time scale back into the Oligocene, about30 Ma ago13. Amazing!

This incredible breakthrough refined the resolution ofthe geological time scale beyond anything previouslyimaginable, except where annual layering was preservedin tree rings, corals, stalagmites or lake sediments, mostof which did not allow dating back beyond about 2000years ago. Palaeoclimatologists who started their careersin the 1980s and later take these advances for granted,but they were astonishing developments to the geologistsof my age group. As Mike Leeder pointed out in 2011,this new understanding was ‘arguably as big an earthsciences discovery as that of plate tectonics’25. MikeWalker and John Lowe agreed, citing the 1976 Hays paperas ‘perhaps the most important Quaternary paper of thepast 50 years’3. It definitively overturned the long-heldperception that there had been four major Quaternaryglacial periods26, by showing that there had been manymore cycles of climate, ice volume and sea level in thelate Cenozoic, and that these cycles formed in response tovariations in Earth’s orbital parameters. As Nick McCaveand Harry Elderfield point out in Shackleton’s obituary,‘This clear recognition of orbital control is also nowrevolutionizing the whole of stratigraphy (the study ofgeological strata) because it provides in principle a meansof correlating beds at separated parts of the Earth to aprecision of 20 000 years at a time of hundreds of millionsof years ago, and of determining precise “orbitally tuned”age-calibrated stratigraphies back to about 250 Maago’27.

This was not all. Given that the climate was gov-erned by celestial mechanics, and that Berger’s dataprojected Earth’s orbital properties and insolation farinto the future, Hays, Imbrie and Shackleton deducedthat ‘the long term trend over the next 20,000 yearsis towards extensive Northern Hemisphere glaciationand a cooler climate’24. That’s not quite the picturewe have today, but it’s close. Berger’s data show thatbecause the Earth’s orbit is at present close to circu-lar, the present warm Holocene interglacial should lastsome 30–50 Ka, of which we have already experienced10 Ka28, 29. So, we should have ∼20 Ka more relativelywarm climate before the next glacial period. For today,Berger’s data show that Earth should be experiencing

a slight cooling trend, which started around 10 Ka ago(Figure 12.1). We look at that more closely in laterchapters. These various dramatic developments wereably summarised by John and Katherine Imbrie in their1979 book Ice Ages: Solving the Mystery18. It remains aclassic.

In the 1980s, the CLIMAP project was succeeded bythe SPECMAP (Spectral Mapping) project, designed byImbrie, Shackleton and others to produce continuous timeseries of Ice Age climate change from deep-sea sedimentsand to facilitate studying their spectral properties. A keySPECMAP achievement was publication of a time scalefor the last 780 Ka, based on a 𝜕

18O reference curvecompiled by stacking together planktonic foraminiferal𝜕

18O records from five low- and middle-latitude sites.Stacking avoids local ‘noise’ interfering with the under-lying signals30. Known as the SPECMAP stack, thiscurve, which was tuned (phase-locked) to the oscilla-tions of precession and obliquity, provided a continuousgeological time scale for the late Pleistocene, the divi-sions of which were accurate to within ±5000 years,an astoundingly high resolution for geological records.The SPECMAP 𝜕

18O stack was improved and extendedover time31, 32. In 2005, Lorraine Lisiecki, then at BrownUniversity, and Maureen Raymo, then at Boston Univer-sity, replaced it with a stack made from combinations ofbenthic foraminiferal data (Figure 12.3)33, 34, which showless variability than planktonic data. Over 100 MISs havebeen identified, going back some 6.6 Ma. The stack is a‘type section’ against which new core measurements arecompared.

The Last Glacial Maximum, or MIS 2, extended from30 to 15 Ka ago. On average, compared with today, it wasthought to be 5–6 ∘C colder globally. It was much colderin the Arctic, with temperatures in central Greenlanddepressed by as much as 20 ∘C35. At that time, much Arc-tic land lay beneath continental ice sheets, and the ArcticOcean was mantled by continuous sea ice and entrappedicebergs. The lack of northward transport of warm, saltywater during the winters made them exceptionally cold.Polar desert replaced tundra. Ice volume peaked about21 Ka ago, after which rising insolation caused ice sheetsand glaciers to melt back, with most coastlines becomingice-free before 13 Ka ago35. We will discuss deglaciationlater.

The climate signal in cores was not smoothly vary-ing, unlike the variation in orbital and axial properties(Figure 12.1). As we can see from Figure 12.3, it wassaw-toothed – something first pointed out by Broecker


3

3.5

4

4.5

3

3.5

4

4.5

5

0 200 400 600 800 1000 1200 1400 1600 1800

51 7

9 1113 15 17 19 21

27

25

29 33

3135

3739

4145

474951 53

55

56

28

20Matuyama Jaramillo

1816

14

1210

8

6

4

2

3

6670

Olduvai 78

1800 2000 2200 2400 2600 2800 3000 3200 3400

3600 3800 4000 4200 4400Time (Ka)

4600 4800 5000 5200

80

7167

7375

7779

9189

8583

81 8793 95

94

97 99 101

102

104

G1G3

G5

G7

G8

G10

G11

Gi1Gi3

2.6

3

3.2

3.4

2.8Gi5

Gi7Gi9

Gi13Gi15

Gi17Gi19

Gi21 Gi25

Gi23

Gi27Co1

Co3CN1 CN5 N1 N5

N7

N6

N9NS1

NS3 Si1Si3

Si5ST1

ST3

T2

T1T3 T5 T7

TG1TG3 TG5

TG4

ThveraSidufjallNunivakCochitiGilbert

NS6

CN7

CN6Gi10

Gi12Gi20

Gi22 Gi26Co2

Co4

G15

G17G19

K1

K2

KM2

KM3

KM6

MG1MG5

MG4

MG6

MG7

MG8

MG11M1

M2

MammothKaenaGauss

G20

G22

Brunhes

57 59 6163

43

23

δ18O

(%

°)δ18

O (

%°)

δ18O

(%

°)

3.6

3.8

Figure 12.3 Benthic oxygen isotope stack, constructed by the graphic correlation of 57 globally distributedbenthic 𝜕

18O records covering 5.3 Ma. Note that the scale of the vertical axis changes from panel to panel. Fromthis stack, a number of new MISs were identified in the early Pliocene. MISs are identified by number back to2.6 Ma ago; before that, the lettering refers to the name of the magnetic chron in which the isotope peaks appear(e.g. Si = Sidufjall, Co = Cochiti etc).

and Van Donk36. The saw-tooth shape represents theslow growth of ice followed by rapid deglaciation, callingfor strong positive-feedback mechanisms to acceleratemelting. We examine this in more detail in Chapter 13.

CLIMAP’s maps of sea surface temperature forthe winter and summer of the Last Glacial Maxi-mum (21.5–18.0 Ka ago) were later updated by theGlacial Atlantic Mapping Project (GLAMAP)37, whichreconstructed the glacial Atlantic Ocean from 275 deep-seasediment cores38. During the northern winter, sea iceextended south as far as about 50∘ N, close to the latitude

of Cork, on the south coast of Ireland. During the northernsummer, warm surface waters moved northwards into theNorwegian-Greenland Sea much as they do today, butthey were between ice sheets on both sides, meeting seaice at about latitude 70∘ N. As today, there was a returnflow from the Arctic down the east side of Greenland. Aproto-Gulf Stream marched eastward from Labrador toLisbon, where the sea surface temperature averaged about16 ∘C.

In parallel with GLAMAP, a comparable project beganto examine Environmental Processes of the Ice Age:


Land, Oceans, Glaciers (EPILOG). It focused on the21 Ka interval, this being the age of minimal summerinsolation at 65 ∘N, and the time when ice sheets reachedtheir maximal volume, as represented by sea level fall39.GLAMAP and EPILOG were followed by the MultiproxyApproach for the Reconstruction of the Glacial Oceanproject (MARGO)40. These various studies and othersconcluded that the last deglaciation began between 22and 18 Ka ago. Two schools of thought emerged, onesuggesting that deglaciation began in the Southern Hemi-sphere, with surface and deep-ocean warming followedby tropical sea surface temperatures and by atmosphericCO2, and the other suggesting that it began in the NorthernHemisphere, where summer insolation at high northernlatitudes was the trigger for ice-sheet decay and sea levelrise41. We explore these issues in Chapter 13.

The MARGO project team constructed maps of sea sur-face temperature to provide constraints on ocean coolingat the Last Glacial Maximum42. Like CLIMAP, they foundthat the strongest mean annual cooling (−10 ∘C) occurredin the mid-latitude North Atlantic, extending into the west-ern Mediterranean, but unlike CLIMAP they found thatthis cooling was most pronounced in the east. Indeed, mostocean basins had cooler eastern than western sides. Thiseastern cooling was probably due to increased upwellingof cold water forced by stronger coastal winds. It was notreplicated by existing GCMs for the Last Glacial Maxi-mum, for reasons not then fully understood (in 2009). Incontrast with CLIMAP, the MARGO team found that con-ditions were ice-free in the summer in the Nordic seas,and that the tropics were on average 1.7 ∘C cooler thanCLIMAP had thought. In the Southern Ocean, the polarfront shifted north from near 60 to 45∘ S, associated with acooling of 2–6 ∘C in the austral winter.

One of the key features evident from the stack of𝜕

18O data (Figure 12.3), noted in 1976 by Shackletonand Opdyke43, is that climate variability has grown withtime19. In the earlier part of the Pleistocene, the signalcomprised relatively small glacial–interglacial changes, inwhich signals of precession (22 Ka cycles) and obliquity(41 Ka cycles) predominated. The signals became muchlarger at about 900 Ka ago, after which a signal with aspacing of 100 Ka intervals predominated. The 100 Kasignal itself became larger with time, especially fromaround 430 Ka ago (MIS 11) onwards. The change at900 Ka ago formed the Mid Pleistocene Transition (MPT).What did it represent?

Harry Elderfield and colleagues used Mg/Ca ratios toestablish what part of the 𝜕

18O signal at the transition

was due to ice volume rather than water temperature44.Changes in ice volume from glacial to interglacial weremuch smaller before the transition than since, presumablybecause the older ice sheets were smaller in area and/orthickness. The transition was a sudden jump, not the resultof a long-term trend towards increased ice volume andcolder temperatures. Elderfield’s team concluded thatit represented ‘an abrupt reorganization of the climatesystem’44. The trigger seemed to be a brief period ofanomalously low summer insolation in the SouthernHemisphere during the warm MIS 23. This suppressedthe melting of ice formed previously in cold MIS 24,allowing unusually extended ice growth in the followingcold MIS 22, at 900 Ka ago, to yield a very large icesheet, associated with a lowering of sea level of about120 m44.

Investigating the behaviours of ice volume and tem-perature, Elderfield’s team confirmed that ice volumefollowed a saw-toothed pattern, growing steadily fromlow amounts during interglacials to high amounts duringglacials, then suddenly retreating. In contrast, bottomwater temperatures followed a square wave pattern,falling to a certain level as ice volume grew, thenstaying more or less constant, before rising again asice volume decreased. The temperatures of bottomwaters during glacial periods remained constant at−1.5–2.0 ∘C, because, once the temperature of surfacewater in the source region fell to about the freezingpoint of salt water, it would fall no further. Bottomwater temperatures warmed to 3 ∘C during interglacialperiods.

Prior to the transition, sea level fell to 70 m below presentlevels. The drop by a further 50 m at the transition exposedcontinental shelf sediments to erosion, transferring marineorganic matter rich in 12C to the deep sea and lowering the𝜕

13C ratio of bottom water and benthic organisms. Elder-field’s team calculated that about half of the fall in 𝜕

13C at900 Ka was due to this change in carbon reservoirs, withthe other half coming from a reduction in the influence ofNorth Atlantic Deep Water44.

As the volume of ice increased across the transition,the supply of aeolian dust, represented by the rates ofaccumulation of sedimentary iron and terrestrial leafwaxes, doubled in the Southern Ocean45. The increase inthe dust supply tells us that the surrounding lands driedout as the globe cooled. The increase in iron helped tocool the globe further, via positive feedback, becausean increase in iron as a key nutrient stimulates produc-tivity, drawing CO2 from the atmosphere, as we see in


more detail later45. This interpretation is supported by anincrease in the sedimentation of opal, representing diatomproductivity, at the same time. The rise in productivitydrove a 30 ppm reduction in CO2 across the transition.By driving a descent into deep, cold, glacial periods,the insolation/dust/CO2 feedback may have initiated thestrong 100 Ka periodicity that characterised subsequentclimate change. On the global scale, an increase in thesupply of dust also coincided with the start of the majorNorthern Hemisphere glaciation at about 2.6 Ma ago,which drew down CO2 in much the same way.

The notion of using a single stack of 𝜕

18O values torepresent Earth’s recent glacial history would have seemedodd to Joseph Adhémar and James Croll, whom we metin Chapter 3, because they thought that cooling related toprecession would alternate between the two hemispheres.As we now know, glaciers and ice sheets in Patagoniaand Antarctica actually advance and retreat at more orless the same times as those in the Northern Hemisphere.Why? The answer lies in those feedbacks to which Crollfirst introduced us. Antarctica is an ice-covered continentsurrounded by ocean – there is nowhere for its land iceto expand into. When Antarctic ice is at a maximum,global ice can only increase by growing on the NorthernHemisphere continents. This growth lowers sea level,exposing Antarctica’s continental shelf and so providingspace for yet more ice growth. Sea level links ice growthon Antarctica to that on the northern continents. Thus,glaciation tends to become more or less synchronous inboth hemispheres, even while the insolation is opposite13.Besides that, insolation has certain seasonal characteristicsthat help to align glaciation in the two hemispheres eventhough their annual insolation signal is opposed, as we seein Chapter 13.

The latest view of the temperature of the Last GlacialMaximum is that it was probably 4.0± 0.8 ∘C cooler thanthe modern preindustrial climate46.

The growing literature on orbital variations and theirrecord in cores from ice, sediments, corals and stalactitesfuelled intensive discussion about the precise mechanismsunderlying the climate changes of the Ice Age, which wereview in Chapter 13.

12.3 The Ice Age CO2 Signal Hiddenon the Deep-Sea Floor

Could carbon isotopes from marine sediments tell us aboutthe abundance of CO2 in the Ice Age atmosphere? Wally

Broecker suggested that the atmospheric CO2 signal couldbe represented by the difference in 𝜕

13C ratios betweensurface planktonic foraminifera and bottom-dwelling ben-thic foraminifera, an idea followed up by Nick Shackletonand colleagues in 198347. The variations they detectedin CO2 by this means matched those found in ice coresby Oeschger, as mentioned in Chapter 9. Clearly, CO2

rose and fell with rises and falls in temperature duringthe Ice Age. How reliable was the association betweenestimated CO2 and the 𝜕

18O values used to estimate tem-perature? In 1985, working with Nick Pisias of OregonState University to analyse the 𝜕

13C and 𝜕

18O profilesthrough the past 340 Ka in a core from 3091 m depth inthe Pacific, Shackleton found that CO2 closely followedtemperature48. It slightly lagged orbital insolation in Juneat 65∘ N, led the response in ice volume (documented byvariations in 𝜕

18O) by about 2500 years and was closelylinked to variations in axial tilt, which dominates theinsolation signal at middle to high latitudes. Shackletonand Pisias concluded that the CO2 signal was forced byhigh-latitude orbital insolation through ‘a mechanismat present not fully understood’48 – probably the effectof that insolation on ocean circulation. As changes inCO2 led changes in ice volume in the North Atlantic,the CO2 must have contributed to the forcing of changesin ice volume there. It was a forcing factor. Insolationwarmed the ocean, which released CO2, which enhancedtemperature, stimulating an eventual decrease in icevolume.

These conclusions were consistent with the proposalby Wally Broecker and Tsung-Hung Peng in their 1982classic Tracers in the Sea that, since the ocean containsabout 60 times more carbon than the atmosphere, theglacial–interglacial change in atmospheric CO2 contentmust have been driven by changes in ocean chemistry49.Broecker seems to have been the first to suggest that aglacial increase in the strength of the biological pumpdrove down CO2 levels. His initial ideas about thebiological pump revolutionised the field of chemicaloceanography50.

Broecker’s ideas got John Martin (1935–1993) thinking.Director of the Moss Landing Marine Laboratories inCalifornia, and crippled by polio when he was 19, Martinset himself the task of figuring out the role of phytoplank-ton – the grass of the sea – in the global climate system51.Phytoplankton use CO2 for photosynthesis. When theirremains sink to the seafloor and decompose, the CO2 isreturned to deep-ocean waters or trapped in sediment andso can no longer contribute to warming the planet. In order


to determine how much plankton sank to the seafloor ina given time, Martin organised the Vertical Transport andExchange of Oceanic Particulate Program (VERTEX) in1981, placing sediment traps across the North Pacific tosample the flux and composition of settling particulates.Among other things, he discovered that the parts of theocean that are high in nutrients but low in chlorophyllwere depleted in iron (Fe)52. Joseph Hart, an Englishscientist, had speculated in the 1930s that this might bethe case, but was unable to prove it. Martin proposed in1990 that Fe was a limiting nutrient and that production ofphytoplankton could be negatively affected by its supply,for example in airborne dust53.

During the Last Glacial Maximum, the supply of dustwas 50 times greater than today, enhancing productiv-ity enough to draw CO2 out of the air. Lack of Fe-richdust during interglacials slowed productivity, leavingCO2 in the air. Martin suggested testing his hypothesisthrough Fe-enrichment experiments even at the scale ofthe whole Southern Ocean. Several such experimentswere carried out, the first in 1993, although none at thescale of an entire ocean. The early experiments found thatthe excess organic matter created by the addition of Fewas recycled in the water column; it did not settle to theseabed as Martin had imagined54. That changed in 2012,when Victor Smetacek of Germany’s Alfred WegenerInstitute for Polar and Marine Research performed a5-week-long Fe-fertilisation experiment in the AntarcticCircumpolar Current with colleagues and discovered thatat least half of the diatom bloom caused by the fertilisationsank far below 1000 m, with much being likely to havereached the deep-sea floor55. This confirmed the viewthat ‘iron-fertilized diatom blooms may sequester carbonfor timescales of centuries in ocean bottom water andfor longer in the sediments’55. Here was support for thegeoengineering notion that iron fertilisation of the oceancould transport CO2 out of the surface waters and into thedeeps, thus drawing down atmospheric CO2.

The abundance of CO2 in the air between glacial andinterglacial times was also governed by the presence orabsence of sea ice56. Growing sea ice placed a lid on polarsurface waters, preventing them from absorbing CO2 fromthe air. As a result, CO2 would gradually accumulate in theair, causing it to warm. Melting of sea ice exposed the coldocean, enabling it to absorb CO2 from the air, contributingto eventual cooling.

12.4 Flip-Flops in the Conveyor

Palaeoceanographic studies radically changed our under-standing of the variability of Ice Age climate and therole of the ocean in climate change. One key resultwas the realisation that the circulation of the ocean haddifferent stable states for glacials and for interglacials.During interglacials like the Holocene, in which welive now, ocean circulation was much as it is today(Figure 12.4).

Warm, salty surface water is drawn towards the Arcticthrough the northern branch of the Gulf Stream, losingheat to the atmosphere en route. This heat warms north-west Europe. By the time the salty water reaches theNorwegian-Greenland Sea, it has cooled to the point ofbecoming dense enough to sink and form North AtlanticDeep Water, which moves south towards Antarcticaand fills the mid-water depths of the Atlantic, Indianand Pacific Oceans. The strong westerly winds blow-ing around Antarctica towards the east force thesenorthern-sourced deep waters to the surface through theprocess of upwelling. The newly upwelled surface watersthen return to the North Atlantic to close the cycle throughtwo pathways. First, under the influence of Antarctica’scoastal easterly winds, some water moves south on tothe Antarctic continental shelf, where the excretion ofsalt from sea ice forming at the ocean’s surface makes itdense enough to sink to the deep-ocean floor. This deep,cold Antarctic Bottom Water moves back to the norththrough the Atlantic, Indian and Pacific Oceans. Becausecold water dissolves larger amounts of oxygen from theatmosphere than does warm water, these deep watersrich in oxygen aerate the bottom of the world’s oceans.Second, much of the rest of the Circumpolar Deep Watereventually wells up to the surface in the Pacific. There,it becomes entrained in the major surface currents thatmove west from the North Pacific through the Indonesianarchipelago, across the Indian Ocean, down the EastAfrican coast in the Agulhas Current and across the SouthAtlantic to the Equator in the Benguela Current, gainingsalt and heat along the way. This water ends up feedingin to the southern end of the Gulf Stream, to repeat thecycle. This global pattern of southward-moving NorthAtlantic Deep Water and northward-moving warm saltysurface water forms the so-called Thermohaline ConveyorBelt (from ‘thermo’, meaning heat, and ‘haline’, meaningsalt), which moves heat and salt around the globe. WallyBroecker is credited with devising this cartoon of oceancirculation57. As we’ll see later, there is a third pathway


not shown in Figure 12.4. Northward-moving surfacewater in the Southern Ocean eventually sinks at the polarfront near 60∘ S to form Antarctic Intermediate Water,which circulates through the world ocean at depths of600–1000 m.

The Thermohaline Conveyor, these days referred to byphysical oceanographers as the Meridional OverturningCirculation (MOC), is geologically young. It did notexist before the opening of the Drake Passage. RobbieToggweiler from Princeton and H. Bjornsson from Icelandused experiments with an ocean model in 2001 to showthat, prior to the opening of the passage, ocean tempera-ture should have been symmetric about the Equator, withmeridional overturning being driven by deep-water forma-tion at the poles in both hemispheres58. With the passageopen, the overturning took the form of an interhemisphericconveyor, with deep-water formation primarily in theNorthern Hemisphere. The conveyor made temperaturesrise in the Northern Hemisphere and fall in the SouthernHemisphere, as the ocean transported heat north across theEquator, especially in the Atlantic. The high salt contentof the warm surface water allowed northern waters tobecome dense when cooled, thus driving the return flowat depth as North Atlantic Deep Water. While salinity dif-ferences are obviously important in driving the conveyor,Toggweiler’s model showed that the conveyor could not be

entirely driven by buoyancy. The westerly winds funnelledthrough Drake Passage do more than ‘set the stage’ for thework of the buoyancy forces in the North Atlantic: they arean indispensible part of the conveyor circulation, becausethey drive the upwelling of deep water around Antarcticato bring North Atlantic Deep Water to the surface58.

In glacial times, the power of the ThermohalineConveyor was much reduced, as we can see from thegeochemistry of microfossils from cores collected atdifferent water depths. In 1982, Ed Boyle from MIT andLloyd Keigwin from Woods Hole Oceanographic Institu-tion (WHOI) found that the shells of benthic foraminiferacontained variations down-core in the ratio of cadmium(Cd) to calcium (Ca)59. The ocean distribution of Cdfollows that of nutrients, so the Cd/Ca ratio in these shellsrecords variations in the nutrient content of bottom waters.North Atlantic Deep Water turned out to be relatively poorin Cd compared with deep water of southern origin. TheCd/Ca evidence told Boyle and Keigwin that the intensityof the northern source relative to the southern one dimin-ished by a factor of two during severe glaciations. 𝜕13Cvalues also document the distribution of nutrients, beinglow in old waters containing abundant dissolved organiccarbon rich in 12C. Both Cd/Ca and 𝜕

13C distributions sug-gest a strong stratification in the North Atlantic at the LastGlacial Maximum, with a low-nutrient, high-𝜕13C water

Arctic Ocean

AtlanticOcean

IndianOcean

PacificOcean

Southern Ocean

Heatreleaseto air

Shallowwarm current

Heatreleaseto air

Heatreleaseto air Deep current

cold and saline

Figure 12.4 Ocean thermohaline conveyor belt, showing the directions and depths of cold, salty, oxygen-richdeep currents, warm surface currents and vertical connections from deep to shallow and vice versa.


mass (Glacial North Atlantic Intermediate Water) occupy-ing depths down to around 2000 m, and a high-nutrient,low-𝜕13C water mass of southern origin below that60.Laurent Labeyrie of the Centre National de la RechercheScientifique (CNRS) laboratory in Gif-sur-Yvette, south-west of Paris, a winner of the European GeophysicalUnion’s Hans Oeschger Medal in 2005, confirmed thatthe overturning circulation of the glacial Atlantic wasshallower and weaker than today’s61. Production of NorthAtlantic Deep Water was much reduced during the LastGlacial Maximum.

With the development of vast ice sheets on westernEurope and North America, the sea’s surface froze duringthe winter as far south, at times, as 40∘ N – the latitudeof Boston in the west and Lisbon in the east. With thefreezing over of the Norwegian-Greenland Sea, there wasno longer a significant source for North Atlantic DeepWater62, 63. Boyle realised that this icing up preventeddeep-water formation64. The Cd/Ca signal showed strongperiodicity at 41 Ka, synchronous with changes in the tiltof the Earth’s axis, confirming that Northern Hemisphereice cover (at least in the Norwegian-Greenland Sea) wascontrolled by insolation at high latitudes. Increasing tiltproduced higher summer insolation and less ice cover.

The icing up of the North Atlantic north of 40∘ N at theLast Glacial Maximum switched off the branch of the GulfStream that extended into the Nordic Seas and deflectedthe Gulf Stream east between New York and Lisbon. Seaice also extended 10∘ closer to the Equator in the SouthernHemisphere, putting a lid on Southern Ocean processes.With the growth of ice on land, sea level fell 120–130 m.The growth of land ice and sea ice increased Earth’s albedosignificantly, helping to cool the planet.

12.5 A Surprise Millennial Signal Emerges

In the 1970s, marine geologists dredged large angularboulders of continental rocks such as granite from theMid-Atlantic Ridge in the North Atlantic, and somepeople unwisely took this to mean that the ridge wasmade of continental rock65. By 1977, Bill Ruddimanof Lamont-Geological Observatory had mapped wideswathes of ice-rafted glacial debris over much of the NorthAtlantic seafloor north of a line connecting Boston withLisbon, and it was realised that the angular boulders werealso ice-rafted66–68. More detailed studies of piston coresfrom the North Atlantic by the German geologist Hartmut

Heinrich (1952–), of the Deutsches HydrographischesInstitut in Hamburg, found ice-rafted debris concentratedin six layers deposited half a precession unit (11 Ka)apart69. Heinrich’s discovery, in 1988, aroused intenseinterest, and an international team formed under the lead-ership of Gerard Clark Bond (1940–2005) of Lamont inorder to investigate these sediments, which Bond’s groupnamed ‘Heinrich layers’70.

The layers formed between 10 and 60 Ka ago, appar-ently in response to surging within the Canadian ice sheet,which led to it breaking up into myriads of icebergs – anarmada carrying rocks across the ocean. Charles Lyellwould have been pleased, since this was the mechanism hehad proposed back in the 1830s to explain the widespreaddistribution of glacial debris across western Europe.However, as we saw in Chapter 2, it is not reasonableto call on drifting icebergs to cover Europe – there, wemust call on a grounded ice sheet to explain the origin ofthe boulders and associated boulder clay. Along with themelting icebergs came vast volumes of cold, fresh water,cooling the North Atlantic. Each event lasted around 750years and began suddenly – within about a decade. Theseoutbreaks reached the continental margin off Portugal,as I discovered in 1995, when I went to sea with NickShackleton on the maiden voyage of the new Frenchresearch vessel Marion Dufresne, with Yves Lancelotas chief scientist. My goal was to collect samples for aproject on ‘Northeast Atlantic Palaeoceanography andClimate Change’ that I had cooked up with geochemistJohn Thomson, from my institute at Wormley in Surrey.We cored the Portuguese continental margin en routefrom the Azores to Marseilles. I can heartily recommendFrench research cruises to those with a taste for cordonbleu cuisine and fine wines.

Much to our delight, Thomson and I found in the MarionDufresne’s 40 m-long giant piston cores certain layers richin magnesium derived from dolomite rock deposited asice-rafted debris by the armadas of melting icebergs fromCanada71, 72. Later, I spent a short sabbatical at Lancelot’slaboratory at Aix-en-Provence, working up some of thedata from our long cores. I remember his kindness inloaning me a car to help me get around during my monththere. It was also a pleasure to meet other prominentFrench palaeoclimatologists during my stay, among themEdith Vincent and Edouard Bard.

The periodic irruption of ice-rafting in Heinrich eventsshowed that the climate of the Ice Age was variable atthe millennial scale, as well as at Milankovitch’s orbital


frequencies. Bond was keen to find out more about thesemillennial-scale processes, and in 1995, with Rusty Lotti,carried out close-spaced analyses down two cores col-lected west of Ireland and spanning the past 9–38 Ka73.Between each of the Heinrich layers they found yet morelayers of ice-rafted debris, but of lesser magnitude. Theydeduced that iceberg calving had recurred at intervals of2000–3000 years. Examination of the rock grains in theselayers showed that, while the carbonate-rich ones camefrom Canada and were concentrated in Heinrich layers,red (hematite)-stained rock grains from multiple sourceswere common, along with grains of volcanic glass fromIceland or Jan Mayen Island, in the other millennial-scalelayers. By 1997, yet more detailed studies by Bondand his team suggested that the red (hematite)-stainedrock grains in these layers came from East Greenlandor Svalbard74. The ice-rafting took place on averageevery 1536± 563 years. This cyclicity persisted intothe Holocene, where the frequency was 1374± 502years – statistically indistinguishable from the cyclicityof the glacial period. Averaging the Holocene and glacialsignals gave a frequency of 1470± 532 years, which, theythought, ‘reflects the presence of a pervasive, at leastquasi-periodic, climate cycle occurring independentlyof the glacial-interglacial climate state’74. Furthermore,‘ocean circulation… [is implicated] as a major factor inforcing the climate signal and in amplifying it during thelast glaciation…with a single mechanism – an oscillatingocean surface circulation, we can explain at once thesynchronous ocean surface coolings, changes in IRD[ice-rafted debris]… and foraminiferal concentrations,and changes in petrologic tracers’74. Laurent Labeyrie’sFrench team distilled this jargon-rich waffle down to theconclusion that a ‘climate oscillator’ caused the millennialoscillations75.

As ever, the science marched on. By 1999, Bond’s teamhad confirmed the persistence of the 1470-year cycle backto 80 Ka ago76. Why weren’t each of these ice-raftingevents associated with massive iceberg outbreaks of thekind forming Heinrich events? Bond thought that aftera massive purge of ice during a Heinrich event, it tooka few thousand years for the ice at source to grow backand reach the unstable conditions needed for anothermassive discharge76. Noting that Heinrich events gotcloser together between MIS 4 and the Last Glacial Max-imum, he thought that this might reflect deterioration ofglobal climate after the last interglacial, with recovery ofthe ice in Hudson Strait taking progressively less time

as the climate cooled. Heinrich events were not strictlyperiodic, and the oscillations were internal to the system.

Bond’s team found signs of oceanic cooling leading upto Heinrich events76, including a southward extension ofpolar surface waters and disruption of the North AtlanticCurrent, the branch of the Gulf Stream that transportswarm waters north75. Mark Maslin of the EnvironmentChange Research Centre at University College Londonsuggested that some of Bond’s 1500-year cooling eventsmay have produced ice surges from Iceland and EastGreenland that were large enough to raise sea level to theextent that it undercut and destabilised the edges of theLaurentide Ice Sheet, precipitating a full-scale Heinrichevent77.

Following up Bond’s work, Bill Curry and his teamfrom the WHOI, on Cape Cod, used 𝜕

18O, 𝜕

13C, andCd/Ca ratios from benthic foraminifera from a core nearIceland to show that these cooling events occurred at moreor less the same time as decreases in the production ofNorth Atlantic Deep Water and cooling of surface watersin the western equatorial Atlantic78. After each Heinrichevent, warm, saline surface water re-entered the area, andthermohaline circulation resumed75, 78. Labeyrie and hisFrench team agreed that millennial ice-rafting events wereassociated with cooling of the surface waters and south-ward extension of cold and low-salinity Arctic waters,but found that the widespread low-salinity meltwateraccompanying these events delayed resumption of theproduction of deep water by several hundred years byforming a lid on the ocean75.

Changes at the millennial scale, such as Heinrich events,Bond cycles and the Younger Dryas cold event at the endof the last glaciation (discussed later), are visible in thereconstructions of past sea surface temperatures madefor the Mediterranean and North Atlantic from alkenonepalaeothermometry79. The Canary Current transports thesemillennial signals south into the tropics along the coast ofnorthwest Africa, where, as a result, the coldest time at thesea surface in the past 80 Ka (−12 ∘C) was not at the LastGlacial Maximum, at around 20 Ka ago, but during Hein-rich event 2, just before the Last Glacial Maximum, andHeinrich event 1, just after the Last Glacial Maximum80.Surface currents also transported these signals, includingall those seen in Greenland ice cores for the past 50 Ka,into the Mediterranean, affecting the climate there, too81.As off northwest Africa, the temperatures of these eventswere colder in the Mediterranean than were those of theLast Glacial Maximum.


Some of the millennial changes were quite fast, andcoincided with large, rapid climate changes recorded inice cores from Greenland. The discovery of these largeand sudden changes in piston cores and ice cores was arevelation, proving that the glacial period was far frombeing as stable as was once supposed. It now appearedthat slow, steady changes in the climate of that time ledeventually to ‘tipping points’ at which the climate changedto a different state, before eventually tipping back to itsprevious one77. Some of the most pronounced changes,especially around the northern North Atlantic, occurredwithin a few decades, or even just a few years. Thesesudden step-like transitions would have had significanteffects on human life at the time, making it prudent for usto reflect on what caused them in the past and what mightdo so in the future as global warming continues. Was this‘flickering’ between one state and another typical justof glacial periods, or might it occur also in interglacialslike the one we are now living through? We revisit thisquestion in Chapters 13 and 14, when we explore theserapid changes in some detail.

Before concluding this section, we should note theconclusion of Julian Dowdeswell, of Scott Polar ResearchInstitute, concerning the behaviour of ice sheet margins.The response of an ice sheet to a climatic event or arise in sea level is not necessarily uniform from oneice margin to another82. Random processes could haveled to surges in Northern Hemisphere ice sheets duringthe last glacial period, which means that even if someexternal forcing agent is involved, there may be a randomcomponent to Heinrich events. Readers who want moredetail on the competing models explaining periodic surg-ing by ice sheets may find it useful to consult Cronin’sPaleoclimates41.

12.6 Ice Age Productivity

Prior to my Marion Dufresne cruise, I had been trying totest the hypothesis that the increase in the steepness of thethermal gradient between the Equator and the poles duringglacial times increased the strength of the Trade Winds andso enhanced upwelling on continental margins. I did so in1995, using a piston core that I had collected back in 1973from the continental slope off Namibia when I was chiefscientist on one leg of rv Chain cruise 115 between Dakarand Cape Town83. I wanted to know whether the upwellingassociated with the Benguela Current changed with time,

and, if so, how and when. The answer required assemblinga multidisciplinary team.

To assess the temperature history, I needed alkenonedata. Fortunately, I knew Geoff Eglinton well, having firstmet him in the late 1970s, when we were both membersof the Organic Geochemistry Panel advising the DSDP.Geoff agreed to provide the alkenone data we needed todetermine sea surface temperatures over the past 70 Ka83.To our surprise, we found that they were coldest andmost productive during MIS 3 (60–24 Ka ago), a warminterstadial during the last glacial period. We deduced thatthe alongshore Trade Winds had been strongest duringstage 3, thus driving more upwelling of cold, nutrient-richand highly productive water. In the colder isotopic stagesabove and below (MIS 2 and MIS 4), waters were slightlywarmer and slightly less productive. While this couldindicate reduced wind strength, the evidence suggestedthat wind directions might have changed, there beingmore winds rich in desert dust blowing directly offshore,and fewer of the alongshore Trades that drove upwellingcurrents. In contrast, today’s sea surface along that marginis very much warmer and less productive than it was inglacial times, although upwelling still prevails there, andsurface waters are still highly productive – this is oneof the world’s great fishing grounds. Our organic carbonsignal fluctuated through time on a cycle of about 22 Ka,evidently driven by variations in the precession of theEarth’s orbit.

Several other researchers were extracting climate sig-nals from piston cores from close by in the southeasternAtlantic at the same time, and we pooled our resourcesto show that the Heinrich events, when icebergs weremost abundant in the North Atlantic, were represented bywarming oceanographic signals in the South Atlantic84.We explore the reasons for this unexpected hemisphericclimatic connection in Chapter 13.

Looking at the Benguela Current system in rather moredetail, a group of German researchers used the alkenonemethod to show that the warming characteristic of the LastGlacial Maximum began before it, probably in responseto a change in the winds that allowed subtropical surfacewaters to move south down the coast from Angola85. Thiscoincided with conditions less favourable for upwelling,which helps to explain the decrease in organic carbonaccumulation we found on the continental slope offWalvis Bay. Timothy Herbert of Brown University foundmuch the same thing off southern California – just a slightcooling at the Last Glacial Maximum, close to the coast. Inboth of these environments, the cores from the open ocean


farther offshore contained temperature profiles typical ofthose seen in the global SPECMAP stack, with the coldestsea surface temperatures at the time of maximum ice vol-ume – the Last Glacial Maximum79. Along the Benguelaand California coast, then, upwelling was stronger andmore productive than during the Holocene during glacials,including at the Last Glacial Maximum, but was not asstrong or productive at the Last Glacial Maximum as itwas in the interstadial period (MIS 3).

Was our finding that upwelling had decreased duringpeak glacial times (MIS 2) typical, I wondered? Yes. AsSigman and Haug explained, the coastal upwelling zonesoff California and Mexico in the north and off Peru inthe south were less productive during the recent glacialperiod50. Sigman and Haug attribute this to the effect ofcontinental cooling (and a large North American ice sheet,in the case of the California Current) on the winds thatcurrently drive coastal upwelling. Upwelling associatedwith monsoonal circulation in the Somali Current of thewestern Indian Ocean also decreased, because the coolingof the Tibetan Plateau weakened the southwest monsoonalwinds. In contrast, upwelling was strengthened in theequatorial Indian Ocean, where the northeast monsoonalwinds remained strong; the same applied in the SouthChina Sea. It was also increased in the eastern equatorialPacific86 and the equatorial Atlantic87.

Yet another geochemical technique helped to ascertainthe history of productivity in the Southern Ocean duringthe Ice Age. Because the element thorium (Th) is rapidlyadsorbed from the ocean on to sinking particles, and thereis little lateral transport of dissolved Th from its site ofproduction to its site of deposition, an isotope of thorium(230Th) can be used as a proxy for the vertical downwardflux of sediment. Use of this technique helped to deter-mine the vertical fluxes of opal, barium, organic carbonand other proxies for palaeoproductivity in the SouthernOcean88. Compared with the Holocene, productivity waslower south of the polar front and higher north of the polarfront during the Last Glacial Maximum in the Atlantic andIndian Ocean sectors51, 90. The main planktonic organismsin these cold waters are siliceous diatoms. Diatom pro-duction shifted north as temperatures cooled. While thisapplied in the Atlantic and Indian Ocean sectors, it did notapply in the Pacific sector, where productivity was lowerin the Last Glacial Maximum than in the Holocene.

The northward shift in productivity reflects northwardmigration of the oceanic fronts and their accompanyingsea ice during glacial times. The absence of high pro-ductivity in the Pacific sector was probably due to its

excessive distance from the westerly sources of dust thattransported Fe to fertilise the ocean. The lower produc-tivity of Antarctic waters during the recent glacial periodwas most likely due to decreasing supply of deep waterto the surface, resulting from the diminished supply ofNorth Atlantic Deep Water, which would have drivena relative fall in atmospheric CO2. In addition, the pre-vailing westerly winds shifted northwards as the HadleyCell shrank, reducing upwelling in the Antarctic coastalsector. Besides that, more extensive cover of sea ice inthe Southern Ocean limited the exposure of the ocean tothe air, contributing to a fall in atmospheric CO2

89–91. Atthe Last Glacial Maximum, sea ice was double its presentextent, both in winter and in summer91. Sigman and Haugsuggest that salinity stratification associated with sea icewas a major limiting factor on CO2 exchange with the airduring glacial times51. Whether there was a net changein total productivity of the Southern Ocean from the LastGlacial Maximum to the Holocene remains a topic fordebate88. It seems more likely that marine productivitystayed the same but underwent a lateral shift from southto north in the Last Glacial Maximum. Regardless of whathappened in the polar regions, studies of 230Th in theequatorial Pacific show little change in productivity fromglacial to interglacial88.

12.7 Observations on Deglaciationand Past Interglacials

The last deglaciation was the most massive change inEarth’s climate in the past 25 Ka. The Northern Hemi-sphere ice sheets began to melt back around 21 Ka ago, asinsolation and CO2 began to rise. Rising seas contributedto the rapid decay of those ice sheets, encouraging anincrease in the rate of flow of ice streams draining theinterior, thus thinning the ice sheets and facilitating theircollapse92. Increased melting formed large meltwaterlakes on the southern fringes of the ice sheets, especiallyin North America, where Lake Agassiz covered an areaabout the size of the Black Sea. Its remnants today formLakes Winnipeg and Manitoba. The sudden drainageof the lake put a freshwater cap on the North Atlantic,shutting down the northern arm of the ThermohalineConveyor. This cap was probably responsible for theYounger Dryas cold period or stadial interrupting thedeglaciation between 12 800 and 11 500 years ago, whichcaused temperatures to drop 5 ∘C in the United Kingdom,for instance. A further drainage from the lake gave rise


to a brief cooling 8200 years ago, which we examine inChapter 14. Other meltwater pulses occurred at around14 200 and 11 000 years ago93. In many respects, the coldYounger Dryas period represented a temporary return tothe glacial circulation pattern of reduced North AtlanticDeep Water. Surprising though it may seem, it was pri-marily a Northern Hemisphere phenomenon, although thealkenone data show that sea surface temperatures fell bysome 12 ∘C off western North America, and demonstratecooling of the same age in the South China Sea, the IndianOcean and the South Atlantic79. The puzzle of how LakeAgassiz drained into the ocean was eventually solved.From gravels and a regional erosion plain in northernCanada, Julian Murton and colleagues at the University ofSussex showed in 2010 that it discharged along the pathof the Mackenzie River94.

Temperatures derived from alkenone data can also beused to check the temperatures obtained for the LastGlacial Maximum by CLIMAP researchers. The alkenonedata show that the surface ocean was cooler then thanCLIMAP researchers thought, but that the tropics cooledmuch less than the high latitudes, perhaps by only 1 ∘C79.

Are past interglacials analogues for the Holocene – theinterglacial we are now living in? The simple answer is: no.Interglacials are not all alike. The modulating effect of theroughly 400 Ka cycle of eccentricity means that the inter-glacial most similar to our own is that from roughly 400 Kaago, during MIS 1195. Analyses of the 𝜕

18O ratios in sam-ples of the right-coiling planktonic foraminifera Neoglobo-quadrina pachyderma from stage 11 in deep-sea drill coresin the northeast Atlantic show that sea surface tempera-tures varied by less than ±1 ∘C from the long-term meanfor at least 30 Ka96. The near-circular orbit of the Earth atthe time prevented the 20 Ka precession signal from hav-ing much effect within this isotope stage. In effect, theMilankovitch cycle ‘missed a beat’, prolonging the inter-glacial to close on 50 Ka. MIS 11 was in effect about twoprecession cycles long, instead of one.

As André Droxler of Rice University in Houston pointedout, MIS 11 and the present interglacial are similar becausetheir orbital variables are almost identical. According toDroxler and colleagues, ‘both interglacials correspond totimes when the eccentricity of the Earth orbit was at itsminimum, so that the amplitude of the precessional cyclewas damped’97. The strongest and longest Pleistoceneinterglacial, stage 11, had prolonged intense warmth, sealevel stands up to perhaps 13–20 m above present levels98

and significant poleward penetration of warm waters. Itlasted twice as long as more recent interglacial stages. The

Holocene is likely to be just as long. In Chapter 13 we willlook at possible explanations for these patterns.

The warming in stage 11 was important for the estab-lishment of coral reefs. Wolf Berger of Scripps and GeroldWefer of the University of Bremen used deep-sea drill coredata to show that the western Pacific warm pool of surfacewater expanded dramatically some 400 Ka ago, helping toexplain the growth of Australia’s Great Barrier Reef99. Atthat time, shallow carbonate platforms grew to the pointwhere they clogged the flow of surface water through theIndonesian islands between the Pacific and Indian Oceans.The warming also triggered the establishment of other bar-rier reefs, like that off Belize97. These reefs grew when thelarge rise in sea level at the end of the previous glacial max-imum extensively flooded fluvial plains, preventing the for-mer supply of riverborne silt and sand from reaching off-shore reef sites.

‘Will such warm conditions be replicated as theHolocene continues?’ asked Droxler and colleagues97.Yes, they concluded, ‘we can expect another ∼20,000years of interglacial conditions, independent of anyanthropogenic forcing’97. Is MIS 11 an exact analoguefor the Holocene? Not according to David Hodell of theUniversity of Florida at Gainsville, who found maximain the 𝜕

13C of the planktonic foraminifera Globigerinabulloides and in fragmented foraminiferal remains in stage11 sediments from a deep-sea drill core from the CapeBasin off South Africa100. The same indicators in sedi-ments from the last 100 Ka and the Holocene have muchlower values, however, and the carbonate compensationdepth was 600 m shallower in stage 11 than it is now.These patterns suggest a lowering in the concentration ofcarbonate ions in the ocean at that time, possibly relatedto the massive building of barrier reefs in shallow waters.Atmospheric and oceanic feedbacks are not operating inexactly the same way today as they were then. Even so, acomparison of sea surface temperatures and of 𝜕18O ratiosfrom benthic foraminifera in the southeast Atlantic showedthat the 11.7 Ka of the Holocene are indeed comparable tothe first 12 Ka of MIS 11101.

What about the last interglacial, the Eemian, duringMIS 5, which began at around 135 Ka ago and lasteduntil around 110 Ka ago? Alkenone and Mg/Ca data frommarine sediments suggest that it was warmer than the lateHolocene by up to 3 ∘C – consistent with stage 5 experi-encing significantly higher orbital insolation79. These newdata improve on the CLIMAP data, which suggested thatthere was little difference between stage 5 and today. Itnow seems likely that stage 5 was as warm as the early


Holocene climatic optimum, when insolation was muchhigher than it is today. We can extract more evidenceof Eemian climate change from pollen and lake recordsin central Europe, loess sediments from central Chinaand marine sediment cores from the eastern subtropicalAtlantic. These data show evidence for a single, suddencool event in the middle Eemian at about 122–120 Kaago, showing that short, sharp cold periods can occur ininterglacials.

Compared with today, global ice volumes were smallerand solar radiation was 13% stronger over the Arcticin summer during the Eemian interglacial. Accordingto Gifford Miller and his team, sea ice and permafrostwere vastly reduced, boreal forest expanded to the Arcticshore and most Northern Hemisphere glaciers melted35.Summer temperature anomalies over Arctic lands were4–5 ∘C above present values, especially in the Atlanticsector. Northern Canada and parts of Greenland were 5 ∘Cwarmer than today in summer, but Alaska and Siberiawere only about 2 ∘C warmer. Interpretation of marinedata is complicated by the stratification of the ArcticOcean, which commonly has a cool relatively fresh cap(<−1 ∘C) overlying warmer subsurface waters (>1 ∘C).

A final point to bear in mind is that interglacials as warmas the present one occurred for only about 10% of the timein the late Quaternary102. The climate of the past 800 Kawas predominantly cold.

12.8 Sea Level

Rising sea level is one of the most highly visible resultsof a warming world, driven by the melting of ice onland and the expansion of warming seawater. Thesechanges are termed ‘eustatic’102. Although changes in sealevel provide us with yet another proxy for past climatechange – especially for ice volume – the relation betweenclimate and sea level is not simple, as we saw in Chapter11. As R. Lawrence Edwards of the University of Min-nesota and his colleagues remind us103, sea level can alsochange as the result of tectonic uplift or sinking of theEarth’s surface, or of isostatic changes, through which landsinks beneath ice sheets but rises around their periphery toform a ‘fore-bulge’ – a process that reverses when the icesheets melt. Only eustatic change is truly global. Tectonicand isostatic adjustments cause local or regional changesthat complicate the extraction of a global sea level signal.Lyell knew all about tectonic effects, having observedthat the so-called Temple at Serapis in Italy had first been

partly drowned and then uplifted. These competing signalsmust be unravelled to separate the local from the globalsignal, as we see in some detail in Chapter 14.

There is also the question of rates. For example, theScandinavian ice sheet melted by about 6000 years ago,but Scandinavia is still slowly rising. So too is Scotland,which lost its ice long ago. In contrast, southern England,the southern edge of the Baltic and the west coasts of Ger-many and the Netherlands, which were on the fore-bulgearound the European ice sheets, are still slowly subsiding.Similarly, the parts of the northernmost United States andCanada that lay beneath the Laurentide Ice Sheet are nowrising, while the southern United States, which formed thefore-bulge area outside that ice sheet, is slowly sinking.Within the area of the former Laurentide Ice Sheet, itscore region – Hudson’s Bay – is still depressed below sealevel, although it is rising slowly.

Past sea levels can be determined directly by using the14C or other radiometric techniques to date carbonatessuch as reefs and other features that formed at or veryclose to sea level103. These techniques include U/Thdating, which involves calculating ages from radioactivedecay relationships between 238U, 234U and 230Th; this isalso known as 230Th dating. A further check on accuracycan be obtained from U/Pa dating, in which ages arecalculated from the relationship between Uranium-235(235U) and its daughter, Protoactinium-231 (231Pa). U/Thand U/Pa dating extend the range of 14C dating (maximum50 Ka) to 250 Ka (230Pa) and 600 Ka (230Th). These tech-niques, like AMS 14C dating, came into their own after themid 1980s, with the development of mass spectrometricmeasurements that reduced sample size and increasedthe speed and precision of analysis. Even so, despite theaccuracy of the dates, all estimates of past sea level comewith some uncertainty.

In this section, we focus on how high sea level mighthave been during past warm interglacials. The data avail-able to Edwards in 2003 suggested that sea levels were upto 20 m above today’s level in MIS 11 (400 Ka ago); up to29 m above in MIS 9 (330 Ka ago); up to 9 m above in MIS7 (240 Ka ago); and around 5± 3 m above in MIS 5, thelast interglacial (100 Ka ago)103. The rise in sea level froma low point of about −130 m following the Last GlacialMaximum is known in some detail, thanks to comparabledata from the New Guinea’s Huon Peninsula, Tahiti, SouthEast Asia’s Sunda Shelf and northwest Australia’s Bona-parte Gulf. These are far-field sites remote from polar icesheets. Isostatic adjustments are unimportant, and the datareflect a true global signal.


These estimates were refined for the last interglacial(MIS 5) by a team led by Robert Kopp of Princeton, whoin 2009 compiled a large number of indicators of local sealevel change and applied a statistical approach to estimat-ing global sea level104. They found a 95% probability thatglobal sea level peaked at least 6.6 m higher than today,and a 67% probability that it exceeded 8 m, but only a 33%likelihood that it exceeded 9.4 m. Rates of sea level risecould have varied between about 56 and 92 cm per century.For comparison, the present rate of sea level rise is around33 cm per century. The last interglacial was only slightlywarmer than the present – by about 2 ∘C. Achieving a sealevel rise in excess of 6.6 m higher than present ‘is likelyto have required major melting of both the Greenland andWest Antarctic ice sheets’, they concluded104.

Eelco Rohling of the University of Southampton useddata from the Red Sea to suggest, like Kopp, that duringthe last interglacial, sea level reached a mean positionof +6 m, with individual short-term peak positions upto about +9 m compared with today’s level. The rate ofrise of sea level, according to Rohling and colleagues,was about 1.6 m per century, which ‘would correspondto disappearance of an ice sheet the size of Greenlandin roughly four centuries’105. This rate occurred whenglobal mean temperature was 2 ∘C higher than today. KurtLambeck (1941–) (Box 12.3) of the Australian NationalUniversity evaluated sea level for the last interglacial byusing tectonically stable sites in the ‘far field’, estimatingthat it was 5.5–9.0 m above today’s106, which is consistentwith the findings of both Kopp and Rohling.

Maureen Raymo of Boston University looked with amember of Kopp’s team, Jerry Mitrovica of Harvard,into the contentious suggestion that Pleistocene shorelinefeatures on the tectonically stable islands of Bermuda and

Box 12.3 Kurt Lambeck.

Lambeck, professor of geophysics at the AustralianNational University in Canberra, was born inUtrecht in the Netherlands. From 2006 to 2010 hewas president of the Australian Academy of Sci-ence. He has been honoured with several awards,among them fellowship in the French and USAcademies of Science and the United Kingdom’sRoyal Society, the international Balzan Price(2012) and the Wollaston Medal of the GeologicalSociety of London (2013).

the Bahamas were more than 20 m higher than today inMIS 11, some 400 Ka ago107. They found both sites to belocated on the outer edge of the peripheral bulge of the Lau-rentide Ice Sheet. To account for postglacial crustal subsi-dence at these sites, their elevations were adjusted down-wards by about 10 m. That reduced eustatic sea level rise to∼6–13 m above the today’s level in the second half of MIS11. This rise was caused by prolonged warmth leading tothe collapse of both the Greenland and West Antarctic IceSheets. Given that the likely maximum rises in sea level forthe melting of the Greenland and West Antarctic Ice Sheetsare 7 and 5 m respectively, the change of 6–13 m suggeststhat changes in the volume of the East Antarctic Ice Sheetwere minor.

Roland Gehrels of the University of Plymouth drewattention to other flaws in the analysis of sea level change,focusing on the rise of sea level since the Last GlacialMaximum108. Fairbanks’ classic paper on sea level rise,published in 1989, and based largely on data from Barba-dos, suggested that sea level was 120 m below present atthe Last Glacial Maximum109. Gehrels cited three possiblesources of error in Fairbanks’ data. First, Barbados liesin an active tectonic setting on the edge of the CaribbeanPlate. Even slight tectonic changes could have affectedthe absolute amount of sea level change registered on theisland. Second, Barbados lay on the trailing edge ofthe glacial fore-bulge pushed up around the margins ofthe Laurentide Ice Sheet, and the collapse of that featurewould have created further vertical change. Third, Fair-banks’ curve included data from other islands (Martinique,Bahamas, Puerto Rico and St Croix), ‘thereby introducingerrors resulting from differential isostatic movements andregional sea-level variations’108.

Focusing on far-field sites, Yokoyama and colleaguescalculated in 2000 that global sea level was as low as130–135 m below present levels at the Last GlacialMaximum110. Claire Waelbroeck and colleagues pro-vided much the same picture, with a maximal loweringto −135 m at the Last Glacial Maximum111. They alsocalculated that sea level fell to about −125 m in MIS 6(140 Ka ago) and MIS 10 (345 Ka ago), and to about−110 m during MIS 8 (250 Ka ago). Their values differsignificantly from those derived by Nick Shackleton112

and provide good reasons for discounting Shackleton’sdata. Lambeck recently estimated sea level lowering atthe Last Glacial Maximum as −134 m113, explaining thatthis was a measure of grounded ice volume, including icegrounded on shelves. Along far-field continental margins,the Last Glacial Maximum sea levels would generally be


less than this due to isostatic/gravitational effects, while inmid oceans they would exceed this (James Scourse, pers.comm.).

When it was published in 1989, the Barbados sea levelcurve109 gained a great deal of attention because it showedevidence for episodes of very rapid sea level rise, mostnotably the event known as meltwater pulse 1a, dated toabout 14 Ka ago, when sea levels rose by 15–25 m at ratesof over 40 mm/yr108. The jury is still out with respect to thesource of the meltwater pulse, which could have originatedin surges of the Laurentide or the Antarctic Ice Sheets, orfrom the discharge of large glacial lakes.

Clearly, sea level has changed through time in responseto the waxing and waning of ice sheets, and measurementsof past sea level can be used as a proxy for ice volumechange. In order to refine these calculations further, thereis much still to learn about regional variations in sealevel, which depend on local tectonics and glacial isostaticadjustments of the Earth’s surface to the addition orremoval of large masses of ice. Knowing that ‘regionalsea-level changes resulting from polar ice melt can departby up to 30% from the global mean’, Gehrels concludedthat ‘regional sea-level variability precludes the use ofthe term “eustasy” in the traditional sense (i.e. globalaverage sea-level change). The recognition that “eustasy”is only a concept should… lead to [improved] regionalsea-level predictions’108.

As we saw in Chapter 11, one way to avoid problemscreated by the use of past shorelines to establish past sealevels is to use the 𝜕

18O composition of seawater, whichis related to both ocean temperature and ice volume, bothof which are, in effect, global. Subtracting the temperaturesignal enables us to determine ice volume and hence sealevel114, 115.

Using stable oxygen isotope analyses of planktonicforaminifera and bulk sediments from the Red Sea, EelcoRohling and his team developed a relative sea level recordfor the past 520 Ka98. It shows a striking similarity tothe record of Antarctic temperature, a relationship thatremains the same regardless of whether the climate systemis shifting towards glaciation or deglaciation, and whichdoes not drift back through time. As this is a robustrelationship within the climate system, it could be appliedin estimating the effects of future climate change (seeChapter 16).

Jacqueline Austermann of Harvard agreed with therevisions to the Fairbanks model of sea level change116.Austermann and colleagues’ model confirmed that at theLast Glacial Maximum, sea level should have been lowered

to about −130 m, not −120 m as Fairbanks thought. Thatleft a significant volume of ice in the Northern Hemisphereunaccounted for: it appeared from sea level data that moreice must have melted than had been available in the icesheets of Laurentia (North America) and Fennoscandia. Ajoint team from Germany’s Alfred Wegener Institute andthe Korean Polar Research Institute discovered in 2013that the furrows that arise when large ice sheets becomegrounded on the seabed are widespread on the seabed offthe coast of northeast Siberia. The team estimated that thefurrows represented the former existence of an Arctic icesheet that covered an area at least as large as Scandinaviaand was up to 1200 m thick. This previously missing icemay well explain the accounting discrepancy117.

The study of sea level is worth an entire book. Read-ers wishing to probe further might like to start with a2010 compilation of global data entitled UnderstandingSea-Level Rise and Variability118.

Returning to the astronomical calculations, it is nowabundantly clear that regular changes in the Earth’s orbitand axial tilt cause the amount of insolation we receiveto vary within narrow limits, a discovery as influential inits own way as plate tectonic theory. The limits define a‘natural envelope’ in which the maxima and minima areseldom if ever exceeded. These limits apply in turn toglobal temperature, which varied over the narrow globalrange of 4–5 ∘C between glacial and interglacial times.Back in 1982, Wolf Berger realised that this was ‘a strik-ing phenomenon, important especially for the survival ofhigher organisms’119. This Ice Age natural envelope wassuperimposed on a background climate whose extremesvaried within another natural envelope, in which, as wesaw at the end of Chapter 9, the variation in CO2 wasdriven by plate tectonic processes including the emissionof CO2 from volcanoes and its extraction by weathering,especially in mountainous areas, and by sedimentationin growing ocean basins. The natural envelope of CO2

from 200 to 1000 ppm (Chapter 9) was only occasion-ally exceeded, as we saw in Chapter 10, when hothouseconditions prevailed.

To summarise, our view of Pleistocene climate changeddramatically from the mid 1960s onwards, when pistoncorers and deep-ocean drilling enabled us to study forthe first time the climate history recorded over the 66%of the Earth’s surface covered by water depths of morethan 200 m. Application of novel palaeontological andgeochemical techniques showed that Earth had experi-enced many more substantial variations in climate thanwas apparent from studies of glaciation on land, where


the advances of later glaciers and ice sheets removed therecords of earlier ones. The realisation that changes ininsolation were intimately linked to changes in tempera-ture and ice volume enabled palaeoclimatologists to tunetheir signals of climate change to orbital changes, thusderiving a novel method for dating core horizons to anunheard of accuracy of ±2000 years, over periods of morethan 1 million years. Furthermore, as clockwork variationsin insolation could be projected into the future, it becamepossible to estimate the extent, duration and timing of thenext glaciation.

The growing global array of deep-ocean cores enabledcomparisons to be made between glacial and interglacialconditions caused by changes in the extent of sea ice andin the Thermohaline Conveyor Belt. This confirmed thevalidity of orbital cycles and highlighted the saw-toothedpattern of actual climate change, reflecting the slowbuild-up of ice sheets and their rapid eventual demise,a pattern suggesting that once warming caused meltingto reach some critical rate, land ice reservoirs collapsedto produce a glacial termination119. Carbon isotopescould be used to estimate the amount of CO2 in the air,showing that CO2 varied with temperature, presumablybecause carbon reservoirs slowly built up in peat beds,rain forest debris, fine-grained organic-rich sediments anddeep-ocean waters as ice accumulated, before decayingrapidly and releasing CO2 as ice melted and the climatewarmed119.

CO2 provided one positive feedback, affecting temper-ature. Sea ice provided another, first through its affect onalbedo, and second through governing the exchange ofCO2 between ocean and atmosphere119. Dust provideda third, increasing fertilisation of the ocean with ironin glacial times, thereby enhancing CO2 draw-down; itsabsence in interglacials had the opposite effect. Watervapour provided a fourth, following CO2, influencingtemperature and governing change in the water cycle. Sealevel provided a fifth, increasing or decreasing the areaof ocean available for the exchange of CO2 and watervapour between ocean and atmosphere. Sea levels in pastinterglacials may have been as high as 9 m above today’s.

A millennial level of natural variability became appar-ent from concentrations of ice-rafted debris. Large glacialoutbreaks formed Heinrich events; small ones formed1500-year Bond cycles. They corresponded with coldperiods in the North Atlantic and warm periods in theSouth Atlantic. Outbreaks from massive glacial lakesflooded the northern oceans from time to time, causing thenorthern freeze known as the Younger Dryas.

In the next chapter, we will look at the exciting discov-eries that the ice core drillers were making on land, andcompare them with those emerging from studies of theocean floor. The history of fossil CO2 in ice cores enablesus to further explore the relationship between CO2 andthe curious 100 Ka climate cycle. We will also examinepossible mechanisms for glacial–interglacial climatechange.

References

1. Milankovitch, M. (1941) Kanon der erdbestrahlung and seineandwendung auf das eiszeitproblem. Special Publication133. Royal Serbian Academy, Belgrade; English translationpublished by Israel Program for Scientific Translations, USDept of Commerce, 1969.

2. Zeuner, F.E. (1959) The Pleistocene Period – Its Climate,Chronology and Faunal Successions, 2nd edn. HutchinsonScientific and Technical, London.

3. Walker, M. and Lowe, J. (2007) Quaternary science 2007: a50-year retrospective. Journal of the Geological Society ofLondon 164, 1073–1092.

4. Weedon, G. (2003) Time-Series Analysis and Cyclostratig-raphy. Cambridge University Press, Cambridge.

5. Berger, A. (1976) Long-term variations of daily and monthlyinsolation during the Last Ice Age. EOS 57 (4), 254.

6. Berger, A. (1976) Obliquity and general precession forthe last 5,000,000 years. Astronomy and Astrophysics 51,127–135.

7. Berger, A. (1977) Support for the astronomical theory of cli-matic change. Nature 268, 44–45.

8. Berger, A. (1978) Long-term variations of caloric insolationresulting from the Earth’s orbital elements. QuaternaryResearch 9, 139–167.

9. Elias, S. (2006) Encyclopedia of Quaternary Science. Else-vier Science, Amsterdam.

10. Arrhenius, G. (1952) Sediment Cores from the East Pacific.Reports of the Swedish Deep Sea Expedition 5. Elandersboktryckeri aktiebolag, Sweden.

11. Ericson, D.B. (1953) Sediments of the Atlantic Ocean. Tech-nical Report on Submarine Geology 1, Lamont GeologicalObservatory, Palisades, New York. Columbia UniversityPress, New York.

12. Emiliani, C. (1955) Pleistocene temperatures. Journal ofGeology 63, 538–578.

13. Hay, W.W. (2013) Experimenting on a Small Planet: A Schol-arly Entertainment. Springer, New York.

14. Emiliani, C. and Flint, R.F. (1963) The Pleistocene record.In: The Sea: Ideas and Observations on progress in the Studyof the Seas, vol. 3, The Earth Beneath the Sea – History (ed.M.N. Hill). John Wiley & Sons, London, pp. 888–927.

15. Revelle, R. and Suess, H.E. (1957) Carbon dioxide exchangebetween atmosphere and ocean and the question of an


increase of atmospheric CO2 during the past decades. Tellus9 (1), 18–27.

16. Shackleton, N.J. and Turner, C. (1967) Correlation betweenmarine and terrestrial Pleistocene successions. Nature 216,1079–1082.

17. Shackleton, N.J. (1967) Oxygen isotope analyses and Pleis-tocene temperatures re-assessed. Nature 215, 15–17.

18. Imbrie, J. and Imbrie, K. (1979) Ice Ages: Solving the Mys-tery. Enslow, Berkeley Heights, NJ.

19. Lea, D.W. (2003) Elemental and isotopic proxies of pastocean temperatures. In: The Oceans and Marine Geochem-istry, Vol. 6, Treatise on Geochemistry (eds H.D. Hollandand K.K. Turekian). Elsevier-Pergamon, Oxford, pp. 365–390.

20. Imbrie, J. and Kipp, N.G. (1969) Quantitative interpretationof late Pleistocene climate based on planktonic foraminiferalassemblages in Atlantic cores (abstract). Geological Societyof America Meeting Program for 1969, part 7, p. 113.

21. Imbrie, J. and Kipp, N.G. (1971) A new micropaleontologi-cal method for quantitative paleoclimatology: application toa late Pleistocene Caribbean core. In: Cenozoic Glacial Ages(ed. K.K. Turekian). Yale University Press, New Haven, CT,pp. 71–181.

22. Imbrie, J., Van Donk, J. and Kipp, N.G. (1973) Paleoclimaticinvestigation of a Caribbean core: comparison of isotopic andfaunal methods. Quaternary Research 3, 10–38.

23. CLIMAP Project Members (1976) The surface of the ice ageEarth. Science 191 (4232), 1131–1137.

24. Hays, J.D., Imbrie, J. and Shackleton, N.J. (1976) Variationsin the earth’s orbit: pacemaker of the ice ages. Science 194(4270), 1121–1132.

25. Leeder, M. (2011) Sedimentology and Sedimentary Basins –From Turbulence to Tectonics, 2nd edn. Wiley-Blackwell,Chichester.

26. Flint, R.F. (1971) Glacial and Quaternary Geology. JohnWiley & Sons, New York.

27. McCave, I.N. and Elderfield, H. (2011) Sir Nicholas JohnShackleton, 23 June 1937–24 January 2006. BiographicalMemoirs of Fellows of the Royal Society 57, pp. 435–462.

28. Berger, A. and Loutre, M.F. (1994) Astronomical forc-ing through geological time. Special Publications of theInternational Association of Sedimentology 19, pp. 15–24.

29. Berger, A. and Loutre, M.F. (2002) An exceptionally longinterglacial ahead? Science 297, 1287–1288.

30. Imbrie, J., Hays, J.D., Martinson, D.G., McIntyre, A.,Morley, J.J., Pisias, N.G., et al. (1984) The orbital theory ofPleistocene climate: support from a revised chronology ofthe marine 𝜕18O record. In: Milankovitch and Climate, Part I(eds A.L. Berger, J. Imbrie, J. Hays, G. Kukla, B. Saltzman).D. Reidel, Dordrecht, pp. 269–305.

31. Imbrie, J., Boyle, E.A., Clemens, S.C., Duffy, A., Howard,W.R., Kukla, G., et al. (1992) On the structure and origin ofmajor glaciation cycles. I. Linear responses to Milankovitchforcing. Paleoceanography 7, 701–738.

32. Imbrie, J., Berger, A., Boyle, E.A., Clemens, S.C., Duffy, A.,Howard, W.R., et al. (1993) On the structure and origin ofmajor glaciation cycles. II. The 100,000-year cycle. Paleo-ceanography 8, 699–735.

33. Lisiecki, L.E. and Raymo, M.E. (2005) A Pliocene-Pleistocene stack of 57 globally distributed benthic 𝛿

18Orecords. Paleoceanography 20, PA1003.

34. http://lorraine-lisiecki.com/stack.html (last accessed 29 Jan-uary 2015).

35. Miller, G.H., Brigham-Grette, J., Alley, R.B., Anderson, L.,Bauch, H.A., Douglas, M.S.V., et al. (2010) Temperatureand precipitation history of the Arctic. Quaternary ScienceReviews 29, 1679–1715.

36. Broecker, W.S. and Van Donk, J. (1970) Insolation changes,ice volumes, and the O-18 record in deep-sea cores. Reviewsin Geophysics and Space Physics 8, 169–197.

37. Pflaumann, U., Sarnthein, M., Chapman, M., d’Abreu, L.,Funnell, B., Huels, M., et al. (2003) Glacial North Atlantic:sea-surface conditions reconstructed by GLAMAP 2000.Paleoceanography 18 (1065), PA000774.

38. Sarnthein, M., Gersonde, R., Niebler, S., Pflaumann,U., Spielhagen, R., Thiede, J., et al. (2003) Overviewof glacial Atlantic Ocean mapping (GLAMAP 2000).Paleoceanography 18 (1030), PA000769.

39. Mix, A.C., Bard, E. and Schneider, R. (2001) Environmen-tal processes of the last ice age: land, oceans, and glaciers(EPILOG). Quaternary Science Reviews 20, 627–657.

40. Kucera, M., Rosell-Melé, A., Schneider, R., Waelbroeck, C.and Weinelt, M. (2005) Multiproxy approach for the recon-struction of the glacial ocean surface (MARGO). QuaternaryScience Reviews 24, 813–819.

41. Cronin, T.M. (2010) Paleoclimates: Understanding ClimateChange Past and Present. Columbia University Press, NewYork.

42. MARGO Project Members (2009) Constraints on the magni-tude and patterns of ocean cooling at the Last Glacial Maxi-mum. Nature Geoscience 2, 127–132.

43. Shackleton, N.J. and Opdyke, N. (1976) Oxygen isotopeand paleomagnetic stratigraphy of Pacific core V28-239late Pliocene to latest Pleistocene. In: Investigation of LateQuaternary Paleoceanography and Paleoclimatology (edsR.M. Cline and J.D. Hays). Geological Society of AmericaMemoirs 145, 449–464.

44. Elderfield, H., Perretti, P., Greaves, M., Crowhurst, S.,McCave, I.N., Hodell, D. and Piotrowski, A.M. (2012)Evolution of ocean temperature and ice volume through themid-Pleistocene climate transition. Science 337, 704–709.

45. Martinez-Garcia, A., Rosell-Melé, A., Jaccard, S.L., Geibert,W., Sigman, D.M. and Haug, G.H. (2011) Southern Oceandust-climate coupling over the past four million years. Nature476, 312–315.

46. Annan, J.D. and Hargreaves, J.C. (2013) A new global recon-struction of temperature changes at the Last Glacial Maxi-mum. Climate of the Past 9, 367–376.


47. Shackleton, N.J., Hall, M.A., Line, J. and Cang, S. (1983)Carbon isotope data in core V19-30 confirm reduced carbondioxide concentration of the ice age atmosphere. Nature 306,319–322.

48. Shackleton, N.J. and Pisias, N.G. (1985) Atmospheric car-bon dioxide, orbital forcing, and climate. In: The CarbonCycle and Atmospheric CO2: Natural Variations Archean toPresent (eds E.T. Sundquist and W.S. Broecker). Geophysi-cal Monographs 32. American Geophysical Union, Washing-ton, DC, pp. 303–317.

49. Broecker, W.S. and Peng, T.H. (1982) Tracers in the Sea.LDGO, Columbia University Press, New York.

50. Sigman, D.M. and Haug, G.H. (2003) Biological pump in thepast. In: The Oceans and Marine Geochemistry, Vol. 6, Trea-tise on Geochemistry (eds H.D. Holland and K.K. Turekian).Elsevier-Pergamon, Oxford, pp. 491–528.

51. http://earthobservatory.nasa.gov/Features/Martin/martin_4.php (last accessed 29 January 2015).

52. Martin, J.H. and Fitzwater, S.E. (1988) Iron deficiency lim-its phytoplankton growth in the North-East Pacific Subarctic.Nature 331 (6154), 341.

53. Martin, J.J. (1990) Glacial-interglacial CO2 change: the ironhypothesis. Paleocanography 5 (1), 1–13.

54. De La Rocha, C.L. (2003) The biological pump. In: TheOceans and Marine Geochemistry, Vol. 6, Treatise onGeochemistry (eds H.D. Holland and K.K. Turekian).Elsevier-Pergamon, Oxford, pp. 83–111.

55. Smetacek, V., Klaas, C., Strass, V.H., Assmy, P., Montresor,M., Cisewski, B., et al. (2012) Deep carbon export from aSouthern Ocean iron-fertilized diatom bloom. Nature 487,313–319.

56. Frakes, L.A., Francis, J.E. and Syktus, J.I. (1992) ClimateModes of the Phanerozoic – The History of the Earth’s Cli-mate over the Past 600 Million Years. Cambridge UniversityPress, Cambridge.

57. Broecker, W.S. (1987) The great ocean conveyor. NaturalHistory Magazine 97, 74–82.

58. Bjornsson, H. and Toggweiler, J.R. (2001) The climatic influ-ence of Drake Passage. In: The Oceans and Rapid ClimateChange: Past, Present and Future (eds D. Seidov, B.J. Hauptand M. Maslin). Geophysical Monographs 126. AmericanGeophysical Union, Washington, DC, pp. 243–259.

59. Boyle, E.A. and Keigwin, L.D. (1982) Deep circulation ofthe North Atlantic over the last 200,000 years: geochemicalevidence. Science 218 (4574), 784–787.

60. Lynch-Stieglitz, J. (2003) Tracers of past ocean circulation.In: The Oceans and Marine Geochemistry, Vol. 6, Treatiseon Geochemistry (eds H.D. Holland and K.K. Turekian).Elsevier-Pergamon, Oxford, pp. 433–451.

61. Labeyrie, L.D. (1992) Changes in the vertical structure ofthe North Atlantic Ocean between glacial and modern times.Quaternary Science Reviews 11, 401–413.

62. McIntyre, A., Kipp, N.G., Bé, A.W.H., Crowley, T., Kellogg,T., Gardner, J.V., et al. (1976) Glacial North Atlantic 18,000

years ago: a CLIMAP reconstruction. Geological Society ofAmerica Memoirs 145, 43–76.

63. CLIMAP (1981) Seasonal reconstructions of the Earth’ssurface at the last glacial maximum. Map Series, TechnicalReport MC-36. Geological Society of America, Boulder,Colorado.

64. Boyle, E.A. (1984) Cadmium in Benthic Foraminifera andAbyssal Hydrography: Evidence for a 41 kyr Obliquity Cycle.Geophysical Monographs 29. American Geophysical Union,Washington, DC, pp. 360–368.

65. Meyerhoff, A.A. and Meyerhoff, H.A. (1972) The newglobal tectonics: major inconsistencies. Bulletin of theAmerican Association of Petroleum Geologists 56 (2),269–336.

66. Ruddiman, W.F. (1977) Late Quaternary deposition ofice-rafted sand in the subpolar North Atlantic (Lat. 40∘ to65∘N). Bulletin of the Geological Society of America 88,1813–1827.

67. Ruddiman, W.F. and McIntyre, A. (1977) Late Quater-nary surface ocean kinematics and climate change inthe high-latitude North Atlantic. Journal of GeophysicalResearch 82 (27), 3877–3887.

68. Ruddiman, W.F. (1977) North Atlantic ice rafting: a majorchange at 75,000 yr B.P. Science 196, 1208–1211.

69. Heinrich, H. (1988) Origin and consequences of cyclicice rafting in the northeast Atlantic Ocean during the past130,000 years. Quaternary Research 29, 142–152.

70. Bond, G.C., Heinrich, H., Broecker, W.S., Labeyrie, L.,McManus, J., Andrews, J., et al. (1992) Evidence for mas-sive discharges of icebergs into the North Atlantic Oceanduring the last glacial period. Nature 360, 245–249.

71. Thomson, J., Nixon, S., Summerhayes, C.P., Schönfeld,J., Zahn, R. and Grootes, P. (1999) Implications forsediment changes on the Iberian margin over the lasttwo glacial/interglacial transitions from (230Th-excess)systematics. Earth and Planetary Science Letters 165,255–270.

72. Thomson, J., Nixon, S., Summerhayes, C.P., Rohling, E.J.,Schönfeld, J., Zahn, R., et al. (2000) Enhanced productivityon the Iberian margin during glacial/interglacial transitionsrevealed by barium and diatoms. Journal of the GeologicalSociety of London 157, 667–677.

73. Bond, G.C. and Lotti, R. (1995) Iceberg discharges intothe North Atlantic on millennial time scales during the lastglaciation. Science 267, 1005–1010.

74. Bond, G., Showers, W., Cheseby, M., Lotti, R., Almasi, P.,DeMonocal, P., et al. (1997) A pervasive millennial-scalecycle in North Atlantic Holocene and glacial climates. Sci-ence 278, 1257–1266.

75. Labeyrie, L., Leclaire, H., Waelbroeck, C., Cortijo, E., Dup-lessy, J.C., Vidal, L., et al. (1999) Temporal variability ofthe surface and deep waters of the north west Atlantic Oceanat orbital and millennial scales. In: Mechanisms of GlobalClimate Change at Millennial Time Scales (eds P.U. Clark,R.S. Webb and L.D. Keigwin). Geophysical Monographs


112. American Geophysical Union, Washington, DC, pp.77–98.

76. Bond, G.C., Showers, W., Elliot, M., Evans, M., Lotti, R.,Hajdas, I., et al. (1999) The North Atlantic’s 1-2kyr climaterhythm: relation to Heinrich Events, Dansgaard/OeschgerCycles and the Little Ice Age. In: Mechanisms of GlobalClimate Change at Millennial Time Scales (eds P.U. Clark,R.S. Webb, L.D. and Keigwin ), Geophysical Monographs112, American Geophysical Union, Washington, DC, pp.35–58.

77. Maslin, M., Seidov, D. and Lowe, J. (2001) Synthesis ofthe nature and causes of rapid climate transitions during theQuaternary. In: The Oceans and Rapid Climate Change:Past, Present and Future (eds D. Seidov, B.J. Haupt andM. Maslin). Geophysical Monographs 126. AmericanGeophysical Union, Washington, DC, pp. 9–52.

78. Curry, W.B., Marchitto, T.M., McManus, J.F., Oppo, D.W.and Laarkamp, K.L. (1999) Millennial-scale changes in ven-tilation of the thermocline, intermediate, and deep waters ofthe glacial North Atlantic. In: Mechanisms of Global ClimateChange at Millennial Time Scales (eds P.U. Clark, R.S. Webband L.D. Keigwin). Geophysical Monographs 112. AmericanGeophysical Union, Washington, DC, pp. 59–76.

79. Herbert, T.D. (2003) Alkenone paleotemperature determina-tions. In: The Oceans and Marine Geochemistry, Vol. 6, Trea-tise on Geochemistry (eds H.D. Holland and K.K. Turekian).Elsevier-Pergamon, Oxford, pp. 391–432.

80. Zhao, M., Beveridge, N.A.S., Shackleton, N.J. and Sarnthein,M. (1995) Molecular stratigraphy of cores off northwestAfrica: sea surface temperature history over the last 80ka.Paleoceanography 10, 661–675.

81. Cacho, I., Grimalt, J.O. and Canals, M. (2002) Responseof the western Mediterranean Sea to rapid climatic vari-ability during the last 50,000 years: a molecular biomarkerapproach. Journal of Marine Systems 33–34 (C), 253–272.

82. Dowdeswell, J.A., Elverhoi, A., Andrews, J.T. and Hebbeln,D. (1999) Asynchronous deposition of ice-rafted layers inthe Nordic seas and North Atlantic Ocean. Nature 400, 348–351.

83. Summerhayes, C.P., Kroon, D., Rosell-Mele, A., Jordan,R.W., Schrader, H.J., Hearn, R., et al. (1995) Variability inthe Benguela Current upwelling system over the past 70,000years. Progress in Oceanography 35, 207–251.

84. Little, M.G., Schneider, R.R., Kroon, D., Price, B., Sum-merhayes, C.P. and Segl, M. (1997) Trade Wind forcing ofupwelling, seasonality, and Heinrich Events as a response tosub-Milankovitch climate variability. Paleoceanography 12(4), 568–576.

85. Kirst, G., Schneider, R.R., Müller, P.J., von Storch, I. andWefer, G. (1999) Late Quaternary temperature variability inthe Benguela Current system derived from alkenones. Qua-ternary Research 52, 92–103.

86. Pedersen, T.F. (1983) Increased productivity in the easternequatorial Pacific during the Last Glacial Maximum (19,000to 14,000 yr B.P.). Geology 11, 16–19.

87. Lyle, M. (1988) Climatically forced organic carbon burial inequatorial Atlantic and Pacific Oceans. Nature 335, 529–532.

88. Anderson, R.F. (2003) Chemical tracers of particle transport.In: The Oceans and Marine Geochemistry, Vol. 6, Treatiseon Geochemistry (eds H.D. Holland and K.K. Turekian).Elsevier-Pergamon, Oxford, pp. 247–291.

89. Crosta, X., Pichon, J.J. and Burckle, L.H. (1998) Applicationof modern analogue technique to marine antarctic diatoms:reconstruction of maximum sea-ice extent at the Last GlacialMaximum. Paleoceanography 13 (3), 284–297.

90. Crosta, X., Sturm, A., Armand, L. and Pichon, J.J. (2004)Late Qaternary sea ice history in the Indian sector of theSouthern Ocean as recorded by diatom assemblages. MarineMicropalaeology 50, 209–223.

91. Gersonde, R., Crosta, X., Abelmann, A. and Armand, L.(2005) Sea-surface temperature and sea ice distribution of theSouthern Ocean at the EPILOG Last Glacial Maximum – aCircum-Antarctic view based on siliceous microfossilrecords. Quaternary Science Reviews 24, 869–896.

92. Denton, G.H. and Hughes, T.J. (1986) The Last Great IceSheets. John Wiley and Sons, New York.

93. Blanchon, P. (2011) Meltwater pulses. In: Encyclopedia ofModern Coral Reefs (ed. D. Hopley). Springer, New York,pp. 683–690.

94. Murton, J.B., Bateman, M.D., Dallimore, S.R., Teller, J.T.and Yang, Z. (2010) Identification of Younger Dryas outburstflood path from Lake Agassiz to the Arctic Ocean. Nature464, 740–743.

95. Berger, A. (2009) Astronomical theory of climate change. In:Encyclopedia of Paleoclimatology and Ancient Environments(ed. V. Gornitz). Springer, Dordecht, pp. 51–57.

96. McManus, J., Oppo, D., Cullen, J. and Healey, S. (2003)Marine isotope stage 11 (MIS 11): analog for Holocene andfuture climate? In: Earth’s Climate and Orbital Eccentricity:The Marine Isotope Stage 11 Question (eds A. Droxler, R.Z.Poore and L.H. Burckle). Geophysical Monographs 137.American Geophysical Union, Washington, DC, pp. 69–85.

97. Droxler, A., Alley, R.B., Howard, W.R., Poore, R.Z. andBurckle, L.H. (2003) Introduction: unique and exceptionallylong interglacial marine isotope stage 11: window intoearth warm future climate. In: Earth’s Climate and OrbitalEccentricity: The Marine Isotope Stage 11 Question (eds A.Droxler, R.Z. Poore and L.H. Burckle). Geophysical Mono-graphs 137. American Geophysical Union, Washington, DC,pp. 1–14.

98. Rohling, E.J., Grant, K., Bolshaw, M., Roberts, A.P., Siddall,M., Hemleben, C. and Kucera, M. (2009) Antarctic temper-ature and global sea level closely coupled over the past fiveglacial cycles. Nature Geoscience 2, 500–504.

99. Berger, W.H. and Wefer, G. (2003) On the dynamics of theice ages: stage 11 paradox, mid-Brunhes climate shift, and100-ky cycle. In: Earth’s Climate and Orbital Eccentricity:


The Marine Isotope Stage 11 Question (eds A. Droxler, R.Z.Poore and L.H. Burckle). Geophysical Monographs 137.American Geophysical Union, Washington, DC, pp. 41–59.

100. Hodell, D.A., Kanfoush, S.L., Venz, K.A., Charles, C.D. andSierro, F.J. (2003) The mid-Brunhes transition in ODP sites1089 and 1090 (subantarctic South Atlantic). In: Earth’sClimate and Orbital Eccentricity: The Marine Isotope Stage11 Question (eds A. Droxler, R.Z. Poore and L.H. Burckle).Geophysical Monographs 137. American GeophysicalUnion, Washington, DC, pp. 113–129.

101. Dickson, A.J., Beer, C.J., Dempsey, C., Maslin, M.A.,Bendle, J.A., McClymont, E.L. and Pancost, R.D. (2009)Oceanic forcing of the Marine Isotope Stage 11 interglacial.Nature Geoscience 2, 428–433.

102. Montañez, I.P., Norris, R.D., Algeo, T., Chandler, M.A.,Johnson, K.R., Kennedy, M.J., et al. (2011) Understand-ing Earth’s Deep Past: Lessons for Our Climate Future.National Academies Press, Washington, DC.

103. Edwards, R.L., Cutler, K.B., Cheng, H. and Gallup, C.D.(2003) Geochemical evidence for Quaternary sea-levelchanges. In: The Oceans and Marine Geochemistry, Vol.6, Treatise on Geochemistry (eds H.D. Holland and K.K.Turekian). Elsevier-Pergamon, Oxford, pp. 343–364.

104. Kopp, R.E., Simons, F.J., Mitrovica, J.X., Maloof, A.C. andOppenheimer, M. (2009) Probabilistic assessment of sealevel during the last interglacial stage. Nature 462, 863–868.

105. Rohling, E.J., Grant, K., Hemleben, C., Siddall, M.,Hoogakker, B.A.A., Bolshaw, M. and Kucera, M. (2008)High rates of sea-level rise during the last interglacial period.Nature Geoscience 1, 38–42.

106. Dutton, A. and Lambeck, K. (2012) Ice volume and sea levelduring the last interglacial. Science 337 (6091), 216–219.

107. Raymo, M.E. and Mitrovica, J.X. (2012) Collapse of polarice sheets during the stage 11 interglacial. Nature 483,453–456.

108. Gehrels, R. (2010) Sea-level changes since the Last GlacialMaximum: an appraisal of the IPCC Fourth AssessmentReport. Journal of Quaternary Science 25 (1), 26–38.

109. Fairbanks, R.G. (1989) A 17,000-year glacio-eustatic sealevel record: influence of glacial melting rates on theYounger Dryas event and deep-ocean circulation. Nature342, 637–642.

110. Yokoyama Y., Lambeck K., de Deckker P., Johnston, P. andFifield, L.K. (2000) Timing of the Last Glacial Maximumfrom observed sea-level minima. Nature 406, 713–716.

111. Waelbroeck, C., Labeyrie, L., Michel, E., Duplessy, J.C.,McManus, J.F., Lambeck, K., et al. (2002) Sea level anddeep water temperature changes derived from benthicforaminifera isotopic records. Quaternary Science Reviews21, 295–305.

112. Shackleton, N.J. (2000) The 100,000-year ice-age cycleidentified and found to lag temperature, carbon dioxide, andorbital eccentricity. Science 289, 1897–1902.

113. Lambeck, K., Yokoyama, Y. and Purcell, A. (2002) Into andout of the Last glacial Maximum: sea level change duringoxygen isotope stages 3-2. Quaternary Science Reviews 21,343–360.

114. Chappell, J. and Shackleton, N.J. (1986) Oxygen isotopesand sea level. Nature 324, 137–140.

115. Elderfield, H. and Ganssen, G. (2000) Past temperature and𝜕

18O of surface ocean waters inferred from foraminiferalMg/Ca ratios. Nature 405, 442–445.

116. Austermann, J., Mitrovica, J., Ltychev, K. and Milne, G.A.(2013) Barbados-based estimate of ice volume at Last GlacialMaximum affected by subducted plate. Nature Geoscience 6,553–557.

117. Niessen, F., Hong, J.K., Hegewald, A., Matthiessen, J., Stein,R., Kim, H., et al. (2013) Repeated Pleistocene glaciation ofthe east Siberian continental margin. Nature Geoscience 6,842–846.

118. Church, J.A., Woodworth, P.L., Aarup, T. and Wilson,W.S. (2010) Understanding Sea-Level Rise and Variability.Wiley-Blackwell, Chichester and Oxford.

119. Berger, W.H. (1982) Climate steps in ocean history – lessonsfrom the Pleistocene. In: Climate In Earth History: Stud-ies in Geophysics (ed W.H. Berger and J.C. Crowell).Report of Panel on Pre-Pleistocene Climates, for the Geo-physics Study Committee of the Geophysics ResearchBoard of the Commission on Physical Sciences, Math-ematics, and Applications of the US National ResearchCouncil, based on a meeting in Toronto in 1980. NationalAcademies Press, Washington, DC, pp. 43–54. (Availablefrom http://www.nap.edu/catalog/11798.html, last accessed29 January 2015).

solving the ice age mystery: the deep-ocean solution

Documents